Arxiv:1203.5589V2 [Math.SG] 5 Oct 2014 Orbit

ON GROWTH RATE AND CONTACT HOMOLOGY

ANNE VAUGON

Abstract. It is a conjecture of Colin and Honda that the number of periodic Reeb orbits of universally tight contact structures on hyperbolic manifolds grows exponentially with the period, and they speculate further that the growth rate of contact homology is polynomial on non-hyperbolic geometries. Along the line of the conjecture, for manifolds with a hyperbolic component that fibers on the circle, we prove that there are infinitely many non-isomorphic contact structures for which the number of periodic Reeb orbits of any non- degenerate Reeb vector field grows exponentially. Our result hinges on the exponential growth of contact homology which we derive as well. We also compute contact homology in some non-hyperbolic cases that exhibit polynomial growth, namely those of universally tight contact structures on a circle bundle non-transverse to the fibers.

1. Introduction and main results The goal of this paper is to study connections between the asymptotic number of periodic Reeb orbits of a 3-dimensional contact manifold and the geometry of the underlying manifold. We first recall some basic definitions of contact geometry. A 1-form α on a 3-manifold M is called a contact form if α ∧ dα is a volume form on M.A (cooriented) contact structure ξ is a plane field defined as the kernel of a contact form. If M is oriented, the contact structure ker(α) is called positive if the 3-form α ∧ dα orients M. The Reeb vector field associated to a contact form α is

the vector field Rα such that ιRα α = 1 and ιRα dα = 0. It strongly depends on α. The Reeb vector field (or the associated contact form) is called non-degenerate if all periodic orbits are non-degenerate (1 is not an eigenvalue of the differential of the first return map). A fundamental step in the classification of contact structures was the definition of tight and overtwisted contact structures given by Eliashberg [21] following Bennequin [2]. A contact structure ξ is said to be overtwisted if there exists an embedded disk tangent to ξ on its boundary. Otherwise ξ is said to be tight. Uni- versally tight contact structures are structures admitting a tight lift on universal cover. A contact form is called hypertight if there is no contractible periodic Reeb arXiv:1203.5589v2 [math.SG] 5 Oct 2014 orbit. Universally tight and hypertight [32] contact structures are always tight. To study a contact structure, it is useful to focus on the periodic orbits of a Reeb vector field. Weinstein conjectured that every contact form on a closed manifold admits a periodic orbit and that was proved in dimension 3 by Taubes [51]. Beyond the existence of a single periodic Reeb orbit, Colin and Honda are interested in the number NL(α) of periodic Reeb orbits with period at most L. They believe it is related to one of the Thurston geometries of the underlying manifold, namely the hyperbolic geometry.

Conjecture 1.1 (Colin-Honda [18, Conjecture 2.10]). For all non-degenerate contact forms α of a universally tight contact structure on a hyperbolic closed 3- manifold, NL(α) exhibits an exponential growth. 1 2 ANNE VAUGON

The main result of this paper is related to Conjecture 1.1 and applies to manifolds with a non-trivial JSJ decomposition including a hyperbolic component that fibers on the circle (see [3] for more information). Theorem 1.2. Let M be a closed oriented connected 3-manifold which can be cut along a nonempty family of incompressible tori into irreducible manifolds including a hyperbolic component that fibers on the circle. Then, M carries an infinite number of non-isomorphic, hypertight, universally tight contact structures such that for all hypertight non-degenerate contact forms α, NL(α) grows exponentially with L. Additionally, if the full contact homology is well defined and invariant and if α is only non-degenerate then NL(α) grows exponentially with L. Currently, contact homology is not defined in full generality. In what follows this assumption will be called Hypothesis H. The proof of the first part of Theorem 1.2 uses only a well-defined contact homology : our assumptions on α assure that the contact homology is always well-defined. However, the second part of the theorem depends on Hypothesis H. See Section 2 for more details. The fibration condition in Theorem 1.2 is not an insurmountable restriction as Agol [1] recently proved the virtually fibered conjecture [52] which says that a hyperbolic 3-manifold fibers on the circle up to finite covering. Note that the situation in Theorem 1.2 is different from the situation in Conjecture 1.1. Contact homology and more generally Symplectic Field Theory (SFT) are invariants of the contact structure introduced by Eliashberg, Givental and Hofer in 2000 [22]. This is a Floer homology invariant. The associated complex is generated by periodic Reeb orbits and the differential “counts” pseudo-holomorphic curves in the symplectization. Besides full contact homology, there exist two simpler contact homologies : cylindrical contact homology [9] and linearized contact homology which depends on a given “augmentation”. Computation of contact homology hinges on finding periodic orbits and and pseudo-holomorphic curve by solving elliptic partial differential equations and this is usually out of reach. The growth rate of contact homology is an invariant derived from the cylindrical or linearized contact homology introduced by Bourgeois and Colin [8]. It describes the asymptotic behavior with L of the number of periodic Reeb orbits with period smaller than L that contribute to contact homology. It is the contact equivalent of the growth rate of symplectic homology introduced by Seidel [50] and used by McLean [48] to distinguish between cotangent bundles and smooth affine varieties. Theorem 1.2 is a corollary of Theorem 1.3. Theorem 1.3. Under the hypothesis of Theorem 1.2, the manifold M carries an infinite number of non-isomorphic, hypertight, universally tight contact structures with an exponential growth rate of cylindrical contact homology restricted to primitive classes. Under Hypothesis H, the growth rate of linearized contact homology is exponential for any pull-back of the trivial augmentation. The “pull-back” augmentations will be explained in Section 2.4. This theorem draws its inspiration in Colin and Honda’s results [18] on exponential growth of contact homology for contact structures adapted to an open book with pseudo- Anosov monodromy. As proved by Thurston [53], a manifold that fibers on the circle is hyperbolic if and only if it is the suspension of a surface by a diffeomorphism homotopic to a pseudo-Anosov map. Colin and Honda speculate further that the growth rate of contact homology is polynomial in the non-hyperbolic situations. These situations are the following. (1) On manifolds with spherical geometry, the growth rate of contact homology for universally tight contact structures is linear. ON GROWTH RATE AND CONTACT HOMOLOGY 3

(2) On manifolds with a geometric structure neither hyperbolic nor spherical, the growth rate of contact homology for universally tight contact structures is usually polynomial. There is however already an exception to the second situation we will soon discuss. In this paper, we study contact structures in a non-hyperbolic situation. We make use of Giroux [29] and Honda [38] classification of positive contact structures on circle bundles to try out Colin and Honda’s conjectures on a broad family of contact structures. Let π : M → S be a circle bundle over a closed compact surface. Figure 1 gives a summary of this classification. Statements such as “tangent to the fibers” or “transverse to the fibers” mean that there exists an isotopic contact structure with this property. Additionally, χ(S) is the Euler characteristic and χ(S, M) the Euler number of the fibration.

universally tight contact structures

non-transverse to the ﬁbers transverse to the ﬁbers

tangent to the ﬁbers non-tangent to the ﬁbers

χ(S) = 0 χ(S) < 0 χ(S, M) < 0 0 ≤ χ(S, M) ≤ −χ(S) quadratic growth exponential growth linear growth generally unknown

Figure 1. Universally tight positive contact structures on circle bundles over a surface S with non-positive Euler characteristic.

In some cases, the contact homology and its growth rate are already known. For instance, contact structures tangent to the fibers are fiberwise covering of (UT S, ξstd) where UTS is the unit tangent bundle over S and ξstd is the contact element contact structure (see for example [27]). In this case, the Reeb flow of the standard contact form associated to a Riemannian metric is the geodesic flow. If the surface is hyperbolic, there exists an unique closed geodesic in each homotopy class [42, Theorem 3.9.5] and the number of homotopy classes has exponential growth with respect to length [49]. Therefore, growth rates of the number of periodic Reeb orbits and of contact homology are exponential. This is an exception to the second statement of the conjecture of Colin and Honda in the non-hyperbolic cases. If S is a torus, universally tight contact structures are standard contact structures on T 3 [28]. The contact homology is known [22] and its growth rate is quadratic, see for example Bourgeois’s Morse-Bott approach to contact homology [4]. Bour- geois also studied contact structures transverse and non-tangent to the fibers with χ(S, M) < 0. He computed contact homology and obtained a linear growth rate. Each of these contact structure has an S1-invariant contact structure in its isotopy class. In this paper we study the other cases where contact structures are universally tight and non-transverse to the fibers. 4 ANNE VAUGON

Definition 1.4 (Giroux [29]). A contact structure ξ on a fiber bundle π : M → S is walled by an oriented multi-curve Γ on S if (1) ξ is transverse to the fibers on M \ π−1(Γ), (2) ξ is transverse to π−1(Γ) and tangent to fibers of π−1(Γ). We call π−1(Γ) a wall. Note that walled contact structures admit an S1-invariant walled contact structure in their isotopy class. The following theorem justifies the previous definition. Theorem 1.5 (Giroux [29, Théorème 4.4]). Universally tight positive contact structures non-transverse to the fibers are exactly contact structures isotopic to a contact structure walled by a non-trivial multi-curve that contains no contractible component. Giroux’s definition of walled contact structures and Theorem 1.5 provide us a way to decompose our manifold into understandable pieces. This brings us to the second half of this paper. Theorem 1.6. Let π : M → S be a circle bundle over a closed compact surface and Sn ξ be a positive contact structure walled by a non-trivial multi-curve Γ = i=0 Γi that contains no contractible component. Let X = M \ π−1(Γ) be the complement of the wall. Denote by X+,...,X+ the connected components of X for which ξ is 1 n+ positively transverse to the fibers and by X−,...,X− those for which ξ is negatively 1 n− transverse to the fibers. Let η be a loop in M. Then, there exists a hypertight contact [η] form α such that the cylindrical contact homology HC∗ (M, α, Q) is well defined and we have the following n± k [η] M ± (1) if [η] = [fiber] for ±k > 0, then HC∗ (M, α, Q) = H∗(Xj , Q), j=1 0 k0 (2) if there exist j ∈ {1, . . . , n} and k > 0 such that [π(η)] = [Γj] , then [η] M 1 HC∗ (M, α, Q) = H∗(S , Q) where I = {i;[π(Γi)] = [π(Γj)]}, I [η] (3) otherwise, HC∗ (M, α, Q) = 0. In addition, if Hypothesis H is satisfied, the growth rate of the contact homology HC(M, ξ, Q) is quadratic. [η] Here HC∗ (M, α, Q) denotes the cylindrical contact homology restricted to the homotopy class [η]. To obtain the whole picture for universally tight positive contact structures on fiber bundles, it remains to compute contact homology of contact structures transverse to the fibers with 0 ≤ χ(S, M) ≤ −χ(S). Although this has not been dealt with yet, it seems reasonable to work out. However, Colin and Honda’s questions remain out of reach as we have yet a lot to learn about contact structures on hyperbolic manifolds. On the other hand, for non-hyperbolic geometries there is already a counterexample as observed above. The following question may provide some alternative way to tackle connections between geometry and periodic Reeb orbits. Question 1. Is the growth rate of contact homology for a given contact structure related to that of the fundamental group of the underlying manifold ? For example, can the growth rate of the fundamental group be an upper bound for the growth rate of contact homology ? (Roughly speaking, in a finitely generated group, the growth rate counts the number of elements that can be written as a product of length n.) Question 2. Are there growth rates of contact homology that lie between quadratic and exponential growths ? ON GROWTH RATE AND CONTACT HOMOLOGY 5

This paper is part of the authors’s the PhD thesis [55]. It is organized as follows. In Section 2, we brieﬂy outline the theory of contact homology. Morse-Bott approach to contact homology is sketched in Section 3. In Section 4, we give a detailed deﬁnition of the growth rate of contact homology. In Section 5 we discuss positivity of intersection for tori foliated by Reeb orbits. We prove Theorem 1.6 in Section 6 and Theorem 1.2 in Section 7.

Acknowledgements. I am deeply grateful to my advisor, Vincent Colin, for his guidance and support. I would also like to thank Frédéric Bourgeois, Paolo Ghiggini, Patrick Massot, François Laudenbach and Chris Wendl for stimulating and helpful discussions. Thanks also go Jean-Claude Sikorav, Jacqui Espina and an anonymous referee for suggesting numerous improvements and corrections and Marc Mezzarobba for proofreading this text. Last, I am grateful for the hospitality of the Unité de Mathématiques Pures et Appliquées (ens Lyon).

2. Contact homology Throughout this paper, we consider only manifolds of dimension 3 and cooriented contact structures. In this section, we give an overview of contact homology over Q which was introduced by Eliashberg, Givental and Hofer [22]. This is an homology build around Gromov’s holomorphic curves [30] which he introduced to the symplectic world in 1985. For more details see [22, 4]. The reader is reminded that the existence and invariance of contact homology in full generality is still in progress. We will elaborate a little bit more within this section, one can also refer to [23]. We start with reviewing some basics. 2.1. Almost-complex structures and holomorphic curves. The symplectization of a contact manifold (M, ξ = ker(α)) is the non-compact symplectic manifold (R × M, d(eτ α) = ω), where τ is the R-coordinate. An almost complex structure on a even-dimensional manifold X is a map J : TX → TX preserving the fibers and such that J 2 = −Id. In addition, on a symplectization, J is adapted to α if ∂ J is τ-invariant, J ∂τ = Rα, Jξ = ξ and J is compatible with ω (i.e. ω(·,J·) is a Riemannian metric). We are interested in J-holomorphic curves. These are curves u : (Σ, j) → (R × M,J) such that du◦j = J ◦du where (Σ, j) is a Riemannian surface. This equation is called the Cauchy-Riemann equation. When J is unspecified we call u a pseudo- holomorphic curve. One can refer to [47] for more information. Theorem 2.1 (see for instance [47, Lemma 2.4.1]). Let U be an open subset of a Riemann surface (S, j) and let (M,J) be a manifold with an almost complex structure. Then, the critical points of any non-constant J-holomorphic map u : (U, j) → (M,J) are isolated. To define contact homology, we consider pseudo-holomorphic maps u :(Σ˙ , j) → R × M where (Σ˙ , j) is a punctured Riemannian surface. For example, the simplest non-constant holomorphic maps are trivial cylinders: if γ is a T -periodic Reeb orbit, the associated trivial cylinder is the map R × S1 −→ R × M (s, t) 7−→ (T s, γ(T t)). Let x be a puncture of Σ˙ and, for some neighborhood of x, choose some polar coordinates (ρ, θ) centered at x. Such a map u = (uR, uM ) is called positively asymptotic to a T -periodic orbit γ in a neighborhood of x if limρ→0 uR(ρ, θ) = +∞ and limρ→0 uM (ρ, θ) = γ (−T θ). Similarly, it is called negatively asymptotic to γ if limρ→0 uR(ρ, θ) = −∞ and limρ→0 uM (ρ, θ) = γ (+T θ). 6 ANNE VAUGON

Holomorphic curves u :(Σ˙ , j) → (R × M,J) with finite Hofer energy are asymptotic to periodic Reeb orbits near each puncture [36, Theorem 1.3]. Recall that the Hofer energy E of u : Σ˙ → R × M is Z ∗ 0 Eα(u) = sup u d(ϕα), ϕ : R → [0, 1], ϕ ≥ 0 . Σ˙ The following proposition is used in Section 5 to prove the smoothness of the projection of a holomorphic curve on M. Proposition 2.2 ([35, Theorem 4.1]). Let (M, ξ = ker(α)) be a contact manifold and J an almost complex structure on (R × M, d(eτ α)) adapted to α. Consider the standard complex structure j on R × S1. For every non-constant map u : (R × S1, j) → (R × M,J) that is not a trivial cylinder, the points (s, t) such that ∂ ∂τ ∈ im(du(s, t)) are isolated. 2.2. Full contact homology. 2.2.1. Periodic orbits. Let γ be a T -periodic Reeb orbit of a contact manifold (M, ξ = ker(α)) and let p ∈ γ. Denote the Reeb flow by ϕt. The linearized return map preserves the contact structure. Its restriction ψT to ξp is a symplecto- morpism of (ξp, dα). Recall that a non-degenerate periodic orbit γ is called even if ψT (p) has two real positive eigenvalues and odd if ψT (p) has two complex conjugate or two real negative eigenvalues. Let γm be the m-th multiple of a simple orbit γ1. Then γm is said to be good if γ1 and γm have the same parity, otherwise, γm is said to be bad. A relative grading of non-degenerate periodic Reeb orbit is given by the Conley- Zehnder index. This is a Maslov type index. Additionally, its parity matches with the definitions of odd and even periodic orbits. We refer to [44] for a precise presentation. 2.2.2. Definition of full contact homology. Let (M, ξ = ker(α)) be a contact manifold with a non-degenerate contact form. The chain complex A∗(M, α) is the free super-commutative unital Q-algebra generated by all good periodic Reeb orbits i.e. the simple periodic orbits and their good multiples. Choose an almost complex structure J adapted to α. To define the differential ∂, consider the moduli 0 0 space M[Z](J, γ, γ1, . . . γn). This is the space of equivalent classes of solutions to Cauchy-Riemann equation that have finite energy, are positively asymptotic to γ, 0 0 negatively asymptotic to γ1 . . . γn and in the relative homotopy class [Z]. By equivalence classes we mean modulo reparametrization of Σ˙ . The R-translation in R×M 0 0 induces a R-action on M[Z](J, γ, γ1, . . . γn) (see [6] for more details). The differ- 0 0 ential counts elements in M[Z](J, γ, γ1, . . . γn)/R when this space is 0-dimensional. To define a homology, we must have ∂ ◦ ∂ = 0. We want to apply Morse homology 0 0 type arguments. However, M[Z](J, γ, γ , . . . γn)/R is not a manifold and we have 0 0 to assume Hypothesis 1. We denote by µ[Z](γ, γ1, . . . γn) the Conley-Zehnder index 0 0 of γ minus the sum of the Conley-Zehnder indices of γ1, . . . γn where all indices are calculated with respect to a trivialization on Z. Hypothesis 1. There exists an abstract perturbation of the Cauchy-Riemann equa- 0 0 tion such that M[Z](J, γ, γ , . . . γn)/R is a union of branched labeled manifolds with 0 0 corners and rational weights, having the expected dimension µ[Z](γ, γ1, . . . γn)+n−2. The problems arise when we take into account multiply covered curves. There are many approaches to proving that the relevant moduli spaces indeed have such a structure as in Hypothesis 1 due to Fukaya and Ono [26], Liu and Tian [45], Hofer, Wysocki and Zehnder [33, 34, 37]. There also exist partial transversality results due to Dragnev [20]. ON GROWTH RATE AND CONTACT HOMOLOGY 7

Let n 0 0 denote the signed weighted counts of points in 0-dimensional com- γ,γ1,...γn 0 0 ponents of M[Z](J, γ, γ , . . . γn)/R for all relative homology classes [Z] [22, 11]. The signs correspond to the orientation of the moduli space and the weights to the multiplicity of the orbits. The diﬀerential of a periodic orbit γ is X nγ,γ0 ,...,γ0 ∂γ = 1 n γ0 . . . γ0 . i ! . . . i !κ(γ0 ) . . . κ(γ0 ) 1 n 0 0 1 l 1 n γ1,...,γn The sum is taken over all the sets of periodic Reeb orbits and we divide by the 0 0 overcount: i1 . . . il are multiplicities in {γ1 . . . γn} and κ(γ) is the multiplicity of γ. The deﬁnition is extended using the graded Leibniz rule. Hypothesis 1 guarantees ∂ ◦∂ = 0. Next, we still want a contact invariant and so we need to change the contact form. We construct a symplectic cobordism between these two forms. Let α1 and α0 be two non-degenerate, homotopic contact forms.

Then there exist c > 0 and a family (αl)l∈R such that (1) lim αl = cα0, l→−∞ (2) lim αl = α1, l→∞ (3) let α denote the 1-form on R × M induced by (αl)l∈R, then dα ∧ dα > 0. Choose almost complex structures J1 and J0 on R × M adapted to α1 and α0. We c c c ∂ Rα denote by J the almost complex structure such that J|ξ = J|ξ and J ∂τ = c . Last, choose an almost complex structure J on R × M compatible with dα and c 0 0 interpolating between J1 and J0 . Let γ be a Rα1 -periodic orbit and γ1, . . . , γn 0 0 be Rcα0 -periodic orbits. Let M[Z](J, γ, γ1, . . . γn) denote the moduli space of J- 0 0 holomorphic curves, positively asymptotic to γ, negatively asymptotic to γ1 . . . γn and in the relative homotopy class [Z]. We have to assume Hypothesis 2 to obtain a nice structure on these moduli spaces. Hypothesis 2. There exists an abstract perturbation of the Cauchy-Riemann equa- 0 0 tion such that M[Z](J, γ, γ1, . . . γn) is a union of branched labeled manifolds with 0 0 corners and rational weights, having the expected dimension µ[Z](γ, γ1, . . . γn)+n−1. Assuming Hypothesis 2, there exists a chain map c ψ((α ,J ), (cα ,J )) : (A (M, α ), ∂ ) → (A (M, cα ), ∂ c ) 1 1 0 0 ∗ 1 J1 ∗ 0 J0 0 0 similar to ∂ and counting elements in M[Z](J, γ, γ1, . . . γn) when this space is 0- dimensional. The induced map in homology c c Ψ((α1,J1), (cα0,J0 )) : HC∗(M, α1,J1) → HC∗(M, cα0,J0 ) does not depend on αl or J. The map Ψ is the key ingredient to obtain invariance of contact homology. Hypothesis H. Hypothesis H is the union of Hypotheses 1 and 2. Theorem 2.3 (Eliashberg-Givental-Hofer). Let (M, ξ) be a closed contact manifold and α be a non-degenerate contact form. Under Hypothesis H, (1) ∂2 = 0, (2) the associated homology HC∗(M, ξ) does not depend on the choice of the contact form, complex structure and abstract perturbation. One can consult [7] for a sketch of proof. If ∂2 = 0 for some contact form α, we denote HC∗(M, α, J) the associated homology. Some computations were car- ried out by Bourgeois and Colin [8] to distinguish toroidal irreducible 3-manifolds, Ustilovsky [54] to prove the existence of exotic contact structures on spheres and Yau [56] who proved that the contact homology of overtwisted contact structures 8 ANNE VAUGON is trivial. Bourgeois [4] provided other computations using Morse-Bott approach to contact homology. Remark 2.4. Under Hypothesis 1, it is reasonable for us to expect the following. Suppose all the images of J-holomorphic buildings positively asymptotic to γ, neg- 0 0 atively asymptotic to γ1 . . . γn are contained in an open set U in R × M. Then U contains the images of all solutions of perturbed Cauchy-Riemann equations that have the same asymptotics for all abstract perturbations. Roughly speaking, holomorphic buildings are glued holomorphic curves and these are deﬁned in [9] in more details. We will only apply this result when the holomorphic buildings are holomorphic cylinders. 2.2.3. Composition of cobordism maps. Consider the special case of proportional contact forms α and cα for c > 0. Let J be an almost complex structure on R × M adapted to α. Then we have the diﬀeomorphism

ϕc : R × M −→ R × M (τ, x) 7−→ (cτ, x) which sends a J-holomorphic curve to a J c-holomorphic curve. The identiﬁcation of geometric periodic Reeb orbits induces an isomorphism

θ(α, J, c):(A∗(M, α), ∂J ) → (A∗(M, cα), ∂J c ). Let

Θ(α, J, c): HC(M, α, ∂J ) → HC∗(M, cα, ∂J c ) denote the induced map on homology. The maps Ψ and Θ have natural composition properties.

Theorem 2.5 (Eliashberg-Givental-Hofer). Let (αi,Ji) be contact forms and adapted almost complex structures on the symplectization of a closed manifold for i = 0, 1, 2. Under Hypothesis H, if there exist cobordisms as deﬁned in Section 2.2.2 between (α2,J2) and (α1,J1) and between (α1,J1) and (α0,J0), then

Ψ((α2,J2), (α0,J0)) = Ψ((α1,J1), (α0,J0)) ◦ Ψ((α2,J2), (α1,J1)).

Proposition 2.6. Let (αi,Ji) be contact forms and adapted almost complex structures on the symplectization of a closed manifold for i = 0, 1. Under Hypothesis H, (1) for all c > 0, we have the following

c c Θ(α0,J0, c) ◦ Ψ((α1,J1), (α0,J0)) = Ψ((cα1,J1 ), (cα0,J0 )) ◦ Θ(α1,J1, c), c (2) if c < 1, one can choose ψ((α0,J0), (cα, J0 )) = θ(α, J0, c).

Sketch of proof. (1) Denote by αl and J the homotopy and almost complex structure used to deﬁne ψ((α1,J1), (α0,J0)). Consider the homotopy cα l and the al- c c most complex structure J = ϕ∗J where ϕ :(τ, x) 7→ (cτ, x). Then ϕ sends J-holomorphic curves to J c-holomorphic curves. (2) Consider the homotopy αl = c(l)α0 between cα0 and α0 where c is a non- decreasing function. Let J0 be an almost complex structure adapted to α0 and C be an anti-derivative of c. The almost complex structure J = ϕ∗J0 where ϕ : (τ, x) 7−→ (C(τ), x) is adapted to αl. Then ϕ sends J0-holomorphic curves to J-holomorphic-curves and the J0-holomorphic curves used to deﬁne

ψ((α0,J0), (α0,J0)) are trivial cylinders. ON GROWTH RATE AND CONTACT HOMOLOGY 9

2.3. Cylindrical contact homology. Let (M, ξ = ker(α)) be a closed contact manifold and assume α is a non-degenerate and hypertight. The chain complex cyl C∗ (M, α) of cylindrical contact homology is the Q-vector space generated by good periodic Reeb orbits associated to the form α. Choose an almost complex structure J on R × M adapted to α. The cylindrical differential ∂cyl counts J-holomorphic 0 rigid cylinders in the moduli space M[Z](J, γ, γ )/R defined in Section 2.2.2. The differential of a periodic orbit γ is

X nγ,γ0 ∂cylγ = γ0 κ(γ0) γ0

0 0 where the sum runs over γ such that γ and γ are of index difference 1 and nγ,γ0 and κ are defined in Section 2.2.2. As we do in the full contact homology version, we assume Hypothesis H for a nice structure on the moduli spaces. Theorem 2.7 (Eliashberg-Givental-Hofer). Let (M, ξ) be a closed contact manifold and α be a non-degenerate hypertight contact form. Under Hypothesis H, (1) (∂cyl)2 = 0; cyl (2) the associated homology HC∗ (M, ξ) does not depend on the choice of an hypertight contact form α, an almost complex structure J and an abstract perturbation. One can consult [7] for a sketch of proof. We may however restrict cylindrical contact homology to a subset of homotopy classes so that Hypothesis H is satisfied. The differential of cylindrical contact homology respects homotopy classes so we can always restrict it. If Λ is a set of free homotopy classes of M we call partial cylin- Λ drical homology restricted to Λ the homology of the chain complex (C∗ (M, α), ∂) Λ where C∗ (M, α) is generated by good periodic Reeb orbits in Λ. If Λ contains only primitive free homotopy classes, it is shown in [20, Corollary 1.9] that for a generic Λ almost complex structure, the partial contact homology HC∗ (M, α, J) is well defined and does not depend on the choice of J or of a hypertight non-degenerate form. Fact 2.8. The morphisms from Theorem 2.5 induce morphisms ψcyl and Ψcyl on the cylindrical contact homology complex and on cylindrical contact homology with similar properties. 2.4. Linearized contact homology. Cylindrical contact homology is a special case of linearized contact homology. Introduced in Chekanov’s work on Legendrian contact homology [13], linearized contact homology was generalized to contact homology by Bourgeois, Ekholm and Eliashberg [9]. Linearization in SFT also appears in Cieliebak and Latschev’s work [14]. One can also refer to [7, 18].

Deﬁnition 2.9. An augmentation ε :(A, ∂) → (Q, 0) is a Q-algebra homomor- phism that is also a chain map. An augmentation ε in (A, ∂) gives a “change of coordinates” a 7→ a = a − ε(a). ε ε ε ε ε Let (A (M, α), ∂ ) denote the new chain complex and write ∂ = ∂1 + ∂2 + ... ε using the ﬁltration by word length. In particular ∂0 = 0. ε 2 Proposition 2.10. If ε is an augmentation, then (∂1) = 0. Let (M, ξ = ker(α)) be a contact manifold with a non-degenerate contact form and ε be an augmentation of A∗(M, α). The linearized contact homology with ε ε ε ε respect to ε, HC (M, α, J), is the homology of (A∗(M, α), ∂1) where A∗(M, α) is the Q-vector space generated by {γ, γ good period orbit}. 10 ANNE VAUGON

Proposition 2.11. Under Hypothesis H, if the contact form α is hypertight, the complex A∗(M, α) admits the trivial augmentation. The linearized contact homology is then the cylindrical contact homology.

Let α0 and α1 be two non-degenerate, homotopic contact forms and

ϕ :(A(M, α1), ∂J1 ) → (A(M, α0), ∂J0 ) be a chain map. If ε0 is an augmentation on (A(M, α0), ∂J0 ) then ϕ induces a pull back augmentation ε1 = ε0 ◦ ϕ on (A(M, α1), ∂J1 ). The morphisms ψ and Ψ described in Theorem 2.5 induce morphisms ψε0 and Ψε0 . We deﬁne θε0 and Θε0 in the same way. Theorem 2.12 (see [7, Theorem 2.8]). Under Hypothesis H, (1) the set of linearized contact homologies

ε {HC (M, α, J), ε augmentation of (A∗(M, α), ∂J )} is an invariant of the isotopy class of the contact structure ξ = ker(α),

(2) let ϕ1, ϕ2 :(A(M, α1), ∂J1 ) → (A(M, α0), ∂J0 ) be two homotopic chain maps

and ε0 be an augmentation on (A(M, α0), ∂J0 ). Let ε1 and ε2 denote the pull-back augmentations by ϕ1 and ϕ2. Then, the map

ε1 ε2 ϕ(ε1, ε2):(A∗ (M, α1), ∂J1 ) −→ (A∗ (M, α1), ∂J1 ) γ − ε1(γ) 7−→ γ − ε2(γ)

induces an isomorphism Φ(ε1, ε2) in homology such that the diagram

Φ(ε1, ε2) ε1 ε2 HC∗ (M, α1,J1) HC∗ (M, α1,J1)

Φ1 Φ2

ε0 HC∗ (M, α0,J0)

commutes where Φ1 and Φ2 are the morphisms induced by ϕ1 and ϕ2.

Augmentations ε1 and ε2 are said to be homotopic, see [7, Section 2.5] for a general deﬁnition.

3. Morse-Bott approach to contact homology Bourgeois introduced Morse-Bott approach to contact homology in his PhD thesis [4] in 2002. It gives a way to compute contact homology when the contact form is degenerate and there exist submanifolds foliated by periodic Reeb orbits. The main idea is to compare the Morse-Bott degenerate situation to non-degenerate situations obtained by perturbing the degenerate form using a Morse function. In this text, we will only use part of the theory on simple examples to compute the contact homology of circle bundles.

3.1. Perturbation of contact forms of Morse-Bott type. Let (M, ξ = ker(α)) be a contact manifold with a contact form α and let ϕt be the Reeb ﬂow. Deﬁnition 3.1. The form α is of Morse-Bott type if (1) the set σ(α) of period of periodic Reeb orbits is discrete, σ(α) is called the action spectrum, (2) if L ∈ σ(α), then NL = {p ∈ M, ϕL(p) = p} is a smooth closed submanifold;

(3) the rank of dα|NL is locally constant and TpNL = ker(dϕL − I). ON GROWTH RATE AND CONTACT HOMOLOGY 11

3 For instance, the standard contact form αn = sin(nx)dy + cos(nx)dz on T is of Morse-Bott type. The Reeb vector ﬁeld is  0 

Rαn =  sin(nx)  cos(nx) and its flow preserves all tori {x = cst}. A torus {x = x0} is foliated by periodic Reeb orbits if and only if sin(nx0) and cos(nx0) are rationally dependent. Another important example is the case of a contact structure transverse to the fibers on a circle bundle and S1-invariant: such a contact structure admits a contact form whose Reeb vector field is tangent to the fibers. The whole manifold is then foliated by periodic Reeb orbits of the same period. 1 The Reeb flow induces an S -action on NL for all L ∈ σ(α). In general, the quotient space SL is an orbifold. However in the examples studied in this paper, the spaces SL will be smooth manifolds. Hence, we assume here that SL is smooth. We now describe how to perturb a contact form α of Morse-Bott type. Fix 0 L ∈ σ(α). For all L ∈ σ(α) ∩ [0,L], choose a Morse function fL0 on SL0 and extend it to NL0 so that dfL0 (Rα) = 0. Then, extend it to M using cut-off functions in such a way that its support is contained in a small neighborhood of NL0 . Let f L denote the sum of all these functions. Perturb the contact form to αλ,L = (1 + λf L)α. Proposition 3.2 (Bourgeois [4, Lemma 2.3]). For all L > 0, there exists λ(L) > 0 such that for all 0 < λ ≤ λ(L), the periodic orbits of αλ,L with period smaller than 0 L correspond to critical points of fL0 on SL0 for L ∈ σ(α) ∩ [0,L]. Additionally, these periodic orbits are non-degenerate.

Remark 3.3. An almost complex structure J on the symplectization R×M which 1 is S -invariant on NL induces a Riemannian metric dα(·,J·) on SL. 3.2. Morse-Bott contact homology. Roughly speaking, the complex of Morse- Bott approach to contact homology is generated by critical points of the functions fL, and the differential counts generalized holomorphic cylinders. Generalized holomorphic cylinders are a combination of holomorphic curves asymptotic to periodic orbits in the spaces NL and gradient lines in the spaces SL. See [4] for more details, [5] for a summary of [4], or [6] for a general presentation. 1 Consider a family of almost complex structures Jλ adapted to αλ,L and S - 0 invariant on NL0 for all L ≤ L. Generalized holomorphic cylinders are limits of Jλ-holomorphic curves as λ → 0 and derive from two main phenomena. On one side, holomorphic buildings appear similarly to the non-degenerate situation: up to reparametrization, a sequence converges in C∞-loc to a holomorphic curve with asymptotic periodic orbits in some intermediate spaces NL. On the other hand, when the asymptotics of two adjacent levels in a holomorphic building differ, pro- jections on SL grow nearer to a gradient trajectory of fL: up to reparametrization, a sequence converges in C∞-loc to a trivial cylinder over any point of the gradient trajectory. The associated compactness theorem derives from Bourgeois’s thesis [4, Chapters 3 and 4]. One can also refer to [10]. In our simpler setting, Bourgeois’s results lead to the following theorems. Theorem 3.4 (Bourgeois [4]). Let π : M → S be a circle bundle over a closed oriented surface carrying an S1-invariant contact form α transverse to the fibers. Fix L > 0 and a Morse-Bott perturbation fL induced by a Morse function f : S → 1 R. Let Jλ be a family of S -invariant almost complex structures on R × M adapted to αλ,L and converging to an almost complex structure J adapted to α as λ → 0. Assume that (f, g) is a Morse-Smale pair where g is the Riemannian metric on S induced by J and α. Fix two critical points x+et x− of f so that index(x+) − 12 ANNE VAUGON index(x−) = 1 and let γ+ and γ− denote the k-th iterates of associated simple periodic Reeb orbits. Then, for all small λ, the moduli space M(γ+, γ−,Jλ)/R is a 0-dimensional manifold. Additionally, M(γ+, γ−,Jλ)/R can be identified with the set of gradient trajectories from x+ to x−, the holomorphic curves are arbitrarily close to cylinders over the gradient trajectories and the orientations induced by contact homology and Morse theory are the same. Theorem 3.5 (Bourgeois [4]). Consider the standard contact form α = sin(x)dy + cos(x)dz 3 on T . Fix L > 0 and a Morse-Bott perturbation fL induced by a Morse function 1 f : S → R with two critical points. Let Jλ be a family of almost complex structures 1 0 on R × M adapted to αλ,L, S -invariant on NL0 for all L ≤ L and converging to 0 an almost complex structure J. Fix L ≤ L and let T be a torus in NL0 . Let γ+ and γ− denote the k-th iterates of two simple periodic orbits in T associated to the critical points of f. Then for all small enough λ, the moduli space M(γ+, γ−,Jλ)/R has exactly two elements with opposite orientations and the holomorphic curves are arbitrarily close to cylinders over gradient trajectories of f. In addition, if γ+ and γ− are not in the same Morse-Bott torus, M(γ+, γ−,Jλ)/R is empty. Remark 3.6. This theorem generalizes to contact forms sin(nx)dy + cos(nx)dz and f(x)dy + g(x)dz if f and g are increasing and decreasing on the same sets as x 7→ sin(nx) and x 7→ cos(nx). These theorems derive from Bourgeois’s work and Hypothesis 1 is always satisfied. Indeed, the solutions of the Cauchy-Riemann equations is the 0-set of a Fred- holm section in a Banach bundle (described in [4, 5.1.1]) and thus a 3-manifold. To achieve transversality of this section, Bourgeois proves that the linearized Cauchy- Riemann operator is surjective on its 0-set by studying its surjectivity for curves close to holomorphic curves (the curves are defined in [4, 5.3.2], the surjectivity is proved in [4, Proposition 4.13 and 5.14]) and then using an implicit function theorem [4, Proposition 5.16]. To obtain the desired moduli space, we quotient the space of solutions by the biholomorphisms of R × S1 and the R-action. The orientation issues are studied in [4, Proposition 7.6]. Corollary 3.7 (Bourgeois [4]). Let M be an oriented circle bundle over a closed oriented surface S carrying an S1-invariant contact structure ξ which is transverse to the fibers. Let f denote the homotopy class of the fiber. Then, for all k > 0, there exists a contact form α such that [f k] HC∗ (M, α, Q) = H∗(S, Q). The cylindrical contact homology is trivial in all other homotopy classes. ∗ Corollary 3.8. Fix n ∈ N . Let αn = sin(nx)dy + cos(nx)dz be the standard 3 contact form on T . Let cy and cz denote the free homotopy classes associated to 1 1 ny nz {0} × {0} × S and {0} × S × {0}. Let η = cy cz be a non-trivial homotopy class. 0 Then there exists a contact form αn such that n [η] 3 0 M 1 HC∗ (T , αn, Q) = H∗(S , Q). i=1 The cylindrical contact homology is trivial in all other homotopy classes.

Note that contact homology distinguishes between the contact structures ker(αn). Following Corollaries 3.7 and 3.8 and assuming Hypothesis 2, we obtain that the growth rate of contact homology is linear in the circle bundle case and quadratic for T 3. In this paper, we will consider the case where the contact manifold is obtained ON GROWTH RATE AND CONTACT HOMOLOGY 13 by gluing together pieces from these two examples. We now turn to the deﬁnition of the growth rate of contact homology.

4. Growth rate of contact homology

4.1. Algebraic setting. The growth rate of a function f : R+ → R+ is said to be n polynomial of order ≤ n if there exists a > 0 such that f(x) ≤ ax for all x ∈ R+. It is said to be exponential if there exist a > 0 and b > 0 such that f(x) ≥ a exp(bx) for all x ∈ R+. More generally, we can define the growth rate type of a function: two non-deceasing functions f : R+ → R+ and g : R+ → R+ have the same growth rate type if there exists C > 0 such that x h ≤ g(x) ≤ h(Cx) C for all x ∈ R+ (see for instance [19]). We call two such function equivalent and denote by Γ(f) the associated equivalence class. With this definition, if f is equivalent to a polynomial of degree n its growth rate is polynomial of order n and if f is equivalent to an exponential, its growth rate is exponential. Note that our definition gives us more precise informations that the growth rate of symplectic homology [48] given by the formula log max(f(x), 1) lim sup x→∞ log(x) and commonly used in other parts of topology [41]. The following algebraic preliminaries are similar to [48]. A filtered directed system is a family of vector spaces (Ex)x∈[0,∞) such that for all x1 ≤ x2, there exists a linear map ϕx1,x2 : Ex1 −→ Ex2 such that

(1) ϕx1,x1 = Id for all x1 ≥ 0,

(2) ϕx1,x3 = ϕx2,x3 ◦ ϕx1,x2 for all 0 ≤ x1 ≤ x2 ≤ x3. A ﬁltered directed system admits a direct limit E = limx→∞ Ex. By deﬁnition, there exist maps ϕx : Ex → E such that the following diagram commutes for all 0 ≤ x1 ≤ x2.

ϕx1,x2 Ex1 Ex2

ϕx1 ϕx2 E

In what follows, we will assume that Ex is a ﬁnite dimensional space for all x ≥ 0.

Deﬁnition 4.1. The growth rate Γ((Ex)) of (Ex) is the growth rate of x 7→ rk(ϕx).

A morphism of ﬁltered directed systems from (Ex)x∈[0,∞) to (Fx)x∈[0,∞) consists of a positive number C and a family of linear maps Φx : Ex −→ FCx such that the following diagram commutes for all 0 ≤ x1 ≤ x2

Φx1 Ex1 FCx1

ϕE ϕF x1,x2 Cx1,Cx2

Φx2 Ex2 FCx2

Two systems (Ex) and (Fx) are isomorphic if there exists a morphism (C, Φ) from 0 (Ex) to (Fx) and a morphism (C , Ψ) from (Fx) to (Ex) such that, for all x ≥ 0, E F ΨCx ◦ Φx = ϕx,CC0x and ΦC0x ◦ Ψx = ϕx,CC0x. 14 ANNE VAUGON

Lemma 4.2. Two isomorphic ﬁltered directed systems have the same growth rate.

Proof. Consider two ﬁltered directed systems (Ex) and (Fx). By deﬁnition, the following diagram

Ex1 lim E ϕx1 u

ϕx,CC0x lim F Id FCx1 ψCx1 v

0 lim E ECC x1

0 commutes. Thus rk(ϕx1 ) ≤ rk(ψCx1 ). Similarly rk(ψx1 ) ≤ rk(ϕC x1 ). 4.2. Action filtration. Let M be a compact manifold and α be a hypertight non- cyl degenerate contact form on M. Fix L > 0 and let C≤L(M, α) be the Q-vector space generated by the good periodic Reeb orbits with period smaller than L. This is a finite dimensional vector space. Under Hypothesis H, since the differen- cyl tial decreases the action, (C≤L(M, α), ∂≤L)L>0 is a chain complex. We denote by cyl (HC≤L(M, α, J))L>0 the associated homology. The inclusion

cyl cyl i : C≤L(M, α) −→ C≤L0 (M, α) induces a linear map in homology for all L0 ≥ L. Similarly, given a set of free Λ homotopy classes Λ, for all L > 0 we deﬁne a chain complex (C≤L(M, α), ∂≤L) and Λ a homology HC≤L(M, α, J).

cyl Λ Fact 4.3. The families (HC≤L(M, α, J))L>0 and (HC≤L(M, α, J))L>0 are ﬁltered directed systems whose morphisms are induced by inclusions. Furthermore

cyl cyl lim HC (M, α, J) = HC∗ (M, α, J) → ≤L Λ Λ lim HC≤L(M, α, J) = HC∗ (M, α, J). → Let M be a compact manifold and α be a non-degenerate contact form on M ε such that (C∗(M, α), ∂) admits an augmentation ε. Then ∂1 decreases the action on Aε(M, α) and we can define a filtered directed system. Definition 4.4. The growth rate of contact homology is the growth rate of the associated filtered directed system.

cyl cyl Remark 4.5. As rk(ϕL) ≤ dim HC≤L(M, α, J) ≤ dim C≤L(M, α), if the growth rate of contact homology is exponential, the number of periodic Reeb orbits grows exponentially with the period. 4.3. Invariance of the growth rate of contact homology. Fact 4.6. The maps from Section 2.2.2 restrict to maps denoted

c ψ≤L((α1,J1), (cα0,J0 )) c Ψ≤L((α1,J1), (cα0,J0 )) ON GROWTH RATE AND CONTACT HOMOLOGY 15 in the ﬁltered case. In addition θ(α, J, c) and Θ(α, J, c) restrict to maps

θ≤L(α, J, c) :(A≤L(M, α), ∂J ) → (A≤cL(M, cα), ∂J c ) c Θ≤L(α, J, c):HC≤L(M, α, J) → HC≤cL(M, cα, J ). Analogous restrictions exist in the cylindrical and linearized situations.

Fact 4.7. Let ϕ1, ϕ2 :(A(M, α), ∂J ) → (A(M, α0), ∂J0 ) be two homotopic chain maps and ε0 be an augmentation on (A(M, α0), ∂J0 ). Let ε1 and ε2 denote the pull-back augmentations by ϕ1 and ϕ2. The map ϕ(ε1, ε2) from Theorem 2.12 induces a map

ε1 ε2 Φ≤L(ε1, ε2): HC≤L(M, α, J) → HC≤L(M, α, J).

Fact 4.8. Let ε0 be an augmentation on (A(M, α), ∂J ) In addition to the properties from Theorem 2.5 and Proposition 2.6, the maps deﬁned in Fact 4.6 satisfy the following properties. (1) For all 0 < c < 1, 1 Θ cα, J c, ◦ Ψ ((α, J), (cα, J c)) ≤L c ≤L

is the map induced by the inclusion HC≤L(α, J) → HC L (α, J). ≤ c c (2) If ϕ1 = ψ((α1,J1), (cα, J )) ◦ ψ((α, J), (α1,J1)) and ϕ2 = θ(α, J, c) then ε2 c 1 ε0 c Φ≤ L (ε2, ε1) ◦ Θ≤L cα, J , ◦ Ψ≤L((α, J), (cα, J )) c c

ε1 ε1 is the morphism induced by the inclusion HC≤L(α, J) → HC L (α, J) ≤ c where ε1 and ε2 denote the pull-back augmentations by ϕ1 and ϕ2.

Proposition 4.9. Let α0 and α1 be two hypertight contact forms on a compact manifold M. Assume that α0 and α1 are homotopic through a family of contact cyl forms. Under Hypothesis H, the two ﬁltered directed systems (HC≤L(M, α0))L≥0 cyl and (HC≤L(M, α1))L≥0 are isomorphic. cyl cyl Proof. The morphisms between HC≤L(M, α1) and HC≤L(M, α0) are cyl cyl ϕL : HC≤L(M, α1) → HC L (M, α0) ≤ c 1 ϕ = Θcyl cα ,J c, ◦ Ψcyl ((α ,J ), (cα ,J c)) L ≤L 0 0 c ≤L 1 1 0 0 and ϕ0 : HCcyl (M, α ) → HCcyl (M, α ) L ≤L 0 ≤ L 1 c0 α 1 1 ϕ0 = Ψcyl 0 ,J c0 , (α ,J ) ◦ Θcyl α ,J , . L ≤ L 0 0 1 1 ≤L 0 0 0 c0 c c These morphisms give an isomorphism by Fact 4.8.

Corollary 4.10. Let α0 and α1 be two hypertight contact forms on a compact manifold M. Assume that α0 and α1 are homotopic through a family of contact forms. Under Hypothesis H, the associated cylindrical contact homologies have the same growth rate.

Proposition 4.11. Let α0 and α1 be two hypertight contact forms on a compact manifold M. Assume that α0 and α1 are homotopic through a family of contact forms. Let Λ be a set of primitive free homotopy classes of M. Then the associated cylindrical partial contact homologies have the same growth rate. 16 ANNE VAUGON

Proof. The restrictions to the primitive classes of the morphisms deﬁned in the proof of Proposition 4.9 give an isomorphism between ﬁltered directed systems. Apply Lemma 4.2 to obtain the desired result.

Proposition 4.12. Let α0 and α1 be two isotopic contact forms, J0 and J1 be two adapted almost complex structures such that (A∗(α0), ∂J0 ) has an augmentation ε0 and ψ((α1,J1), (α0,J0)) exists. Let ε1 be the pull-back augmentation of ε0 (see Section 2.4). Then, under Hypothesis H, the two ﬁltered directed systems ε1 ε0 (HC≤L(α1,J1))L≥0 and (HC≤L(α0,J0))L≥0 are isomorphic. Thus, the growth rates of linearized contact homology are the same. Proof. Consider the morphisms

ε1 ε0 ϕL : HC≤L(M, α1,J1) → HC≤L(M, α0,J0) ε0 ϕL = Ψ≤L ((α1,J1), (α0,J0)) and

0 ε0 ε1 ϕL : HC≤L(M, α0,J0) → HC L (M, α1,J1) ≤ c 1 c 0 ε1 α0 c c 0 ε0 1 ϕ = Ψ ,J , (α ,J ) ◦ Φ L (ε , ε ) ◦ Θ α ,J , L ≤ L 0 1 1 ≤ 0 0 ≤L 0 0 c c c c 1 c 1 c 0 where ε0 is the pull back augmentation of ε0 by θ c α0,J0 , c and ε0 is the pull 1 1 c back by ψ((α1,J1), (α0,J0))◦ψ c α0,J0 , c , (α1,J1) . These morphisms give an isomorphism by Fact 4.8. 5. Positivity of intersection and tori foliated by Reeb orbits Introduced by Gromov [30] and McDuﬀ [46], positivity of intersection states that, in dimension 4, two distinct pseudo-holomorphic curves C and C0 have a ﬁnite number of intersection points and that each of these points contributes positively to the algebraic intersection number C · C0. In this text, we will only consider the simplest form of positivity of intersection: let M be a 4-dimensional manifold, C and C0 be two J-holomorphic curves and p ∈ M so that C and C0 intersect 0 0 transversely at p. Consider v ∈ TpC and v ∈ TpC two non-zero tangent vectors. 0 0 Then (v, Jv, v , Jv ) is an oriented basis of TpM (J orients TpM). In the contact world, positivity of intersection results in the following lemma. Lemma 5.1. Let (M, ξ) be a contact manifold, α be a contact form and J be an adapted almost complex structure on the symplectization. Consider U an open subset of C, u : U → R × M a J-holomorphic curve and z ∈ U such that duM (z) is injective and transverse to Rα(uM (z)). Then, Rα(uM (z)) is positively transverse to duM (z).

Proof. Let γ :[−ε, ε] → M be an arc in a Reeb trajectory such that γ(0) = uM (z). Consider the holomorphic curve v : R × [−ε, ε] −→ R × M (s, t) 7−→ (s + uR(z), γ(t)). ∂ The holomorphic curves u and v intersect transversely at u(z) and ∂τ ,Rα(uM (z)) is an oriented basis for the tangent plane to v at u(z). The projection of u to M is smooth as duM (z) is injective. Positivity of intersection gives the desired result.

The hypothesis “duM (z) injective and transverse to Rα(uM (z))” is generic (see Theorem 2.1 and Proposition 2.2). We will use positivity of intersection in the following situation. Let (M, ξ = ker(α)) be a contact manifold with a chart I ×S1 × S1 where I is an interval and coordinates (x, y, z) such that α = f(x)dy + g(x)dz. ON GROWTH RATE AND CONTACT HOMOLOGY 17

1 1 Assume that, for all x ∈ I, the tori {x} × S × S = Tx are incompressible in M. A torus T is incompressible in M if the map π1(T ) → π1(M) is injective. Consider u : R × S1 → R × M a pseudo-holomorphic cylinder with ﬁnite energy and asymptotic to the periodic Reeb orbits γ+ and γ− in M. Assume that u intersects R × Tx for all x ∈ I.

Lemma 5.2. There exists a nonempty open interval I1 ⊂ I such that for all x0 ∈ I1 −1 1 u u(R × S ) ∩ (R × Tx0 ) is a disjoint union of smooth circles homotopic to {∗} × S1.

1 1 Proof. There exists a nonempty open interval I1 ⊂ I such that I1 × S × S does −1 1 1 1 not intersect γ+ and γ−. Thus u u(R × S ) ∩ I × S × S is contained in a 1 ∂ compact subset of R × S . As the points such that du(s, t) = 0 or ∂τ ∈ im(du(s, t)) 1 1 1 are isolated in R × S , we may assume that I1 × S × S does not contain images ∂ of points such that du(s, t) = 0 or ∂τ ∈ im(du(s, t)). 1 ∂ Consider x0 ∈ I1 and (s, t) ∈ R × S such that u(s, t) ∈ R × Tx0 . As ∂τ ∈/ im(du(s, t)) and u is pseudo-holomorphic, Rα(s, t) ∈/ im(du(s, t)) and ∂ Span ,R (u(s, t)) ∩ im(du(s, t)) = {0}. ∂τ α As du(s, t) 6= 0, it holds that ∂ Span ,R (u(s, t)) ⊕ im(du(s, t)) = T ( × M). ∂τ α u(s,t) R

Thus im(du) + T (R × Tx0 ) = Tu(R × M) and, by transversality, −1 1 u u(R × S ) ∩ (R × Tx0 ) is a 1-dimensional compact submanifold of R × S1. −1 1 By contradiction, if u u(R × S ) ∩ (R × Tx0 ) has a contractible component

C, then u(C) = c is contractible in R × M. As c ⊂ R × Tx0 and Tx0 is an ∂ incompressible torus, c is contractible in R × Tx0 . As Span ∂τ ,Rα(u(s, t)) ∩ im(du(s, t)) = {0}, the projection of c to M is smooth and transverse to Rα. Yet −1 1 the torus Tx0 is foliated by Reeb orbits. Thus u u(R × S ) ∩ (R × Tx0 ) has only non-contractible components and, as it is a smooth manifold, these components are 1 homotopic to {∗} × S .

Let C be a circle given by Lemma 5.2, then C inherits the orientation of {∗}×S1 and induces a homotopy class of Tx0 . Let p be a vector tangent to Tx0 so that the straight line in Tx0 directed by p is in the homotopy class [C]. If A is a collar neighborhood of C, denote by A± the two connected components of A \ C corresponding to the connected component of R×S1 \C asymptotic to {±∞}×S1.

Lemma 5.3. If (p, Rα) is an oriented basis of Tx0 and A is small enough, then there exist x− and x+ in I1 such that 1 1 1 1 uM (A−) ⊂ (x−, x0) × S × S and uM (A+) ⊂ (x0, x+) × S × S . Otherwise

1 1 1 1 uM (A−) ⊂ (x0, x+) × S × S and uM (A+) ⊂ (x−, x0) × S × S . In other words, holomorphic cylinders cross a torus foliated by periodic Reeb orbits in just one direction. 18 ANNE VAUGON

Proof. Let C(t) be a parametrization of C and c be the projection of u(C) on Tx0 , 0 c is a smooth curve transverse to Rα. If c (t0),Rα(c(t0)) is an oriented basis of 0 Tx0 for some t0, then c (t),Rα(c(t)) is an oriented basis for all t. Thus (p, Rα) is 0 an oriented basis Tx0 if and only if c (t),Rα(c(t)) is an oriented basis. 1 1 The sets uM (A±) are connected and therefore contained in (x0, x+) × S × S or 1 1 0 in (x−, x0)×S ×S . Let V be a normal vector to C at C(t) so that (V,C (t)) is an 0 oriented basis (V points toward A+). Consider v = duM (C(t))·V , then (v, c (t),Rα) is an oriented basis by positivity of intersection. If (p, Rα) is an oriented basis then the x component of v is positive. Conversely, if (Rα, p) is an oriented basis, the x component of v is negative.

6. Contact homology of walled contact structures In this section we prove Theorem 1.6. The strategy of the proof is to decompose the manifold and its walled contact structure into understandable pieces and to study the contact homology of the pieces. Theorem 1.6 states that the cylindrical contact homology of a walled contact structure is the sum of the cylindrical contact homologies of the components of this decomposition. Let π : M → S be a circle bundle over a closed oriented surface and ξ be a contact structure on M walled by a curve Γ that contains no contractible components. To prove Theorem 1.6 we first construct in Section 6.1 a contact form “almost Morse- Bott” such that (1) in a neighborhood of the wall that is a union of thickened tori (one for each connected component of the wall), Reeb orbits foliate all the tori; (2) elsewhere, the Reeb vector field is tangent to the fibers.

This contact form is not of Morse-Bott type as some spaces NT have a nonempty boundary. We study the associated periodic Reeb orbits in Section 6.2. Then, in Section 6.3, we perturb the contact form from Section 6.1 as in the More-Bott case and control periodic Reeb orbits. We prove the quadratic growth rate of contact homology. In Section 6.4, we prove that there are no holomorphic cylinders between two components of the decomposition using positivity of intersection and end the proof using Morse-Bott theory. Remark 6.1. Contact structures ker(sin(nx)dy+cos(nx)dz) on T 3 = T 2 ×S1 with π kπ 1 coordinates (x, y, z) are walled by the curves {( 2n + n , y), y ∈ S }. Theorem 1.6 and Corollary 3.8 give the same contact homology. 6.1. “Almost Morse-Bott” contact form. The following proposition results from Giroux’s work [29]. Proposition 6.2. Let ξ be a contact structure on a circle bundle M walled by a nonempty multi-curve Γ that contains no contractible component. Then there exist a contact structure ξ0 isotopic to ξ, a defining contact form α for ξ0, and a n 1 1 neighborhood U of the wall with local coordinates (x, y, z) in (−1, 1) × ∪i=1S × S such that: −1 n 1 1 (1) π (Γ) '{0} × ∪i=1S × S , (2) ξ0 is walled by Γ, (3) on a trivialization S0 × S1 of M \ U, we have α = β + edz where β is a 1-form on S0 and e = 1 when ξ is positively transverse to the fiber and e = −1 when ξ is negatively transverse to the fiber, (4) α = f(x)dy + g(x)dz on U where f :[−1, 1] → R is negative and strictly convex, and g :[−1, 1] → R has an inflection point at 0, g = −1 on 1 1 [−1, − 2 ], g = 1 on [ 2 , 1] and g is increasing in between, ON GROWTH RATE AND CONTACT HOMOLOGY 19

(5) the change of coordinates between U and a neighborhood of M \U is a linear map (x, y, z) 7→ (x, y, z + ky).

1 1 − 2 2

f 0 f g g0

Figure 2. Maps f and g

∂ Remark 6.3. On M \ U, the Reeb vector ﬁeld is ± ∂z . On U,  0  1 R = −g0(x) . α f 0(x)g(x) − g0(x)f(x)   f 0(x) The open set U is a union of thickened tori foliated by Reeb orbits.

n 1 1 −1 Proof. Let (−1, 1) × ∪i=1S × S be a chart of a neighborhood U of π (Γ) ' n 1 1 ∂ 1 {0} × ∪i=1S × S with coordinates (x, y, z) so that ∂x ∈ ξ and S is the ﬁber. In this chart, any contact form is written α = f(x, y, z)dy + g(x, y, z)dz where g(0, y, z) = 0 and g(x, y, z) 6= 0 for all x 6= 0. Orient Γ so that Γ is negatively transverse to ξ. Without loss of generality, one can assume f(0, y, z) = −1. Consider the path of contact forms αs = sf(x, y, z) + (1 − s)f(x, 0, 0) dy + sg(x, y, z) + (1 − s)g(x, 0, 0) dz

n 1 1 in a small neighborhood of {0} × ∪i=1S × S . For all s ∈ [0, 1], αs is a contact form as f(0, y, z) = f(0, 0, 0) and g(0, y, z) = g(0, 0, 0). Using Moser’s trick, we can n 1 1 find a vector field Xs near {0} × ∪i=1S × S such that (1) Xs(0, y, z) = 0, ∂ (2) Xs is collinear to ∂x , ∗ (3) ker(ϕsα) = ker(αs) where ϕs is the flow of Xs. Extend Xs to M × [0, 1] using a cut-off function. Then ϕs is well defined for all ∗ s ∈ [0, 1]. The contact structure associated to ϕsα is transverse to the fibers on −1 ∗ M \π (Γ) as ϕsα = (f ◦ϕs)dy +(g ◦ϕs)dz, (ϕs)|π−1(Γ) = Id and α = αs on M \U. Therefore ξ is isotopic to a contact structure with a defining contact form α and a 0 n 1 1 −1 chart U = (−1, 1)×∪i=1S ×S near π (Γ) where α = f(x)dy +g(x)dz, g(0) = 0 and g(x) 6= 0 for all x 6= 0. By the contact condition, it holds that g0(0) > 0 and 1 1 one can assume that g = −1 on [−1, 2 ] and g = 1 on [ 2 , 1]. For each connected component of Γ, choose f0 and g0 so that

(1) f0 is negative and strictly convex, 1 (2) g0 :[−1, 1] → R has an inﬂection point at 0, g0 = −1 on [−1, − 2 ], g0 = 1 1 on [ 2 , 1] and g0 is increasing in between, (3) f0 = f near x = ±1, 20 ANNE VAUGON

(4) f0(x) (resp. g0(x)) is the same for all connected components and for all 3 3 x ∈ − 4 , 4 . Write f(x) + ig(x) = ρ(x) exp(iθ(x)) and f0(x) + ig0(x) = ρ0(x) exp(iθ0(x)). By the contact condition, θ and θ0 are decreasing and have the same image as g(x) = 0 (resp. g0(x) = 0) if and only if x = 0. By Gray’s stability theorem on the path (1 − s)ρ(x) + sρ0(x) exp i((1 − s)θ(x) + sθ0(x))

0 we obtain an isotopic contact form such that α = f0(x)dy+g0(x)dz on U . Consider 1 1 1 1 Ui ⊂ U the neighborhood of the component Γi with coordinates (− 2 , 2 ) × S × S . n Let V be a neighborhood of a connected component of M \ ∪i=1Ui. As Γ 6= ∅ and S is connected, V is a manifold with boundary and the circle bundle is trivial. Let S0 × S1 be a trivialization such that the change of coordinates between V and n 1 1 (−1, 1) × ∪i=1S × S is linear (i.e. (x, y, z) 7→ (x, y, z + ky)) in polar coordinates near the boundary. Therefore α = β + edz near ∂V . On V , α = βz + hdz and h 6= 0, so one can assume h = e. By use of Gray’s theorem on the path αs = sβz(x) + (1 − s)β0(x) + edz we obtain the desired contact form. We are already in position to prove the third part of Theorem 1.6.

Proof of Theorem 1.6, part (3). Let η be a loop that neither is in the homotopy class of a multiple of the ﬁber nor projects to a multiple of a curve Γi. Then, the Reeb vector ﬁeld associated to the contact form α given in Proposition 6.2 does no have periodic orbit in the homotopy class [η]. Therefore, there exists a contact [η] form α such that HC∗ (M, α) = 0.

6.2. Rα periodic orbits. We ﬁrst investigate the growth rate of Morse-Bott tori. Let α be a contact form given in Proposition 6.2. Consider Ui ⊂ U the neighborhood 1 1 1 1 n of the component Γi with coordinates (− 2 , 2 ) × S × S . Let W = M \ ∪i=1Ui (see Figure 3). On W , all the ﬁbers are periodic orbits of period 1. On Ui, α =

(y, z) U

Ui W

x 1 1 −

Figure 3. Sets U, Ui and W f(x)dy + g(x)dz and the Reeb vector ﬁeld is given by  0  1 R = −g0 . α f 0g − fg0   f 0

0 0 0 0 g (x) p There are two cases: f (x) 6= 0 and g (x) 6= 0. If f (x) 6= 0 and − f 0(x) = q with 1 1 gcd(p, q) = 1, the period of the periodic Reeb orbits in Tx = {x} × S × S is  0 0  (f g − fg )q  T =   = |qg + fp| .  f 0  ON GROWTH RATE AND CONTACT HOMOLOGY 21

0 0 f (x) q If g (x) 6= 0 and − g0(x) = p with gcd(p, q) = 1, the period of the periodic Reeb orbits in Tx is  0 0  (f g − fg )p T =   = |qg + fp| .  g0  In what follows we will assume q ≥ 0. Recall that we write σ(α) the action spectrum, NL = {p ∈ M, ϕL(p) = p} and SL = NL/S1 for all L ∈ σ(α). Lemma 6.4. The set σ(α) is discrete and #(σ(α)∩[0,L]) exhibits exactly quadratic growth with L. Proof. We first show that it has a quadratic upper bound. There exist A > 0 and intervals I1 and I2 such that 1 1 (1) I1 ∪ I2 = (− , ), 2 2 (2) 1 < A on I and 1 < A on I , g0 1 f 0 2 1 0 0 0 0 (3) A < f g − fg < A, |f | < A and |g | < A. 0 0 Let L > 0. There are at most 3A6L2 rational numbers q such that (f g−fg )p < L p g0(x) f 0(x) q f 0(x) for some x ∈ I1 with − g0(x) = p . As x 7→ − g0(x) is increasing, for all rational q f 0(x) q numbers p there is one x such that −g0(x) = p . Therefore the growth rate is at most quadratic. We now show that the growth rate is also at least quadratic: consider B > 0 0 2 2 f (x) q and p, q such that p + q ≤ B. Then, there exists x such that − g0(x) = p . The 2 associated torus is foliated by periodic Reeb orbits of period smaller than A B. 6.3. Periodic orbits of the perturbed contact form. Let α be a contact form given in Proposition 6.2. We use the notation from Section 6.2. Similar to the method we use in the Morse-Bott case, we perturb the degenerate contact form. 0 Fix L > 0. For all L ≤ L, fix a Morse function fL0 on SL0 and extend it to NL0 0 so that dfL0 (Rα) = 0. We assume that for L = 1, f1 does not depend on y and ∂f1 e ∂x > 0 in the cylindrical coordinates (x, y, z) near ∂W . Extend fL0 to M by using 0 cut-off functions in a standard way. Let f L denote the sum of fL0 for all L ≤ L and perturb the contact form with f L. More precisely, let

αλ,L = (1 + λf L)α.

Note that the diameter in the x-coordinate of connected components of supp(f L) that do not contain W tends to 0 as L → ∞. Additionally, the ﬂow of RαL,λ preserves supp(f L) and M \ supp(f L). Lemma 6.5. For all L > 0, there exists λ(L) > 0 such that for all 0 < λ ≤ λ(L), the periodic orbits of αλ,L with period smaller than L correspond to critical points 0 of fL0 on SL0 for L ∈ σ(α) ∩ [0,L]. These periodic orbits are non-degenerate. Proof. Outside a neighborhood of ∂W , Morse-Bott theory applies directly (see [4, Lemma 2.3]). In a neighborhood of ∂W , in the trivializing chart of W with coordinates (x, y, z), the contact form is written

α = (f(x) + kg(x))dy + g(x)dz = fW (x)dy + g(x)dz as the change of coordinates is linear (Proposition 6.2). As f L only depends on x, we have  0  1 0 R = −g0(1 + λf ) − λgf . αL,λ 0 0 2  L L  (fW g − fW g )(1 + λf L) 0 0 fW (1 + λf L) + λfW f L 22 ANNE VAUGON

In a small neighborhood of ∂W and for λ small enough, the y coordinate of the Reeb vector ﬁeld is as small as desired and does not vanish. Therefore there is no periodic Reeb orbit with period smaller than L. Lemma 6.6. Let η be a loop that is a multiple of the ﬁber or projects to a multiple of a connected component of Γ. Then, there exist L0 > 0 and L 7→ λ(L) positive and decreasing such that for all L ≥ L0 and λ ≤ λ(L)

(1) the periodic Reeb orbits of αL,λ homotopic to η have a period smaller than L, (2) αL,λ is hypertight. In addition, there exist arbitrarily small non-degenerate and hypertight perturbations of αL,λ.

Proof. Let Ry and Rz denote the y and z-coordinates of the Reeb vector ﬁeld. Let W1,...,Wm be the connected components of W . Consider m n [ 0 [ 0 Wi ∪ Uj i=1 j=1 0 −1 0 an open covering of M such that Wi ∩π (Γ) = ∅ for all i = 1 . . . m and Uj ∩W = ∅ for all j = 1 . . . n (see Figure 4). There exists ε > 0 such that, in the trivialization

(y, z) Ui W

Uj0

Wi0 x

0 0 Figure 4. Sets Uj and Wi

0 0 of Wi induced by Proposition 6.2, |Rz| > ε and, in the trivialization of Uj induced 0 by Proposition 6.2, |Ry| > ε. If there exists a loop η in Wi (resp. Uj) such that 0 0 [η ] = [η], let ki (resp. kj) denote the multiplicity of the fiber (resp. Γj) in the 0 0 decomposition of η in the associated trivialization. Else, set ki = 1 (resp. kj = 1). Consider L0 > 0 such that 0 1 (1) L0 > max ({|ki|, |kj|}) · , 1≤i≤m,1≤j≤n ε (2) periodic orbits of Rα homotopic to η have period smaller than L0, 00 (3) for all L ≥ L0 , the connected components of supp(fL00 ) are contained 0 0 either in a Uj or in a Wi . Let L ≥ L0. By Lemma 6.5, there exists λ(L) such that for all λ ≤ λ(L), all periodic Reeb orbits of αL,λ with period smaller than L are non-degenerate and such that 0 ε is a lower bound for the z-component of RαL,λ in Wi and for the y-component of 0 RαL,λ in Uj. Let γ be a αL,λ-periodic Reeb orbit with period greater than L. Then, either γ ⊂ supp(fL) or γ ⊂ M \ supp(fL). If γ ⊂ M \ supp(fL) then γ is not homotopic 0 to η by condition (2). If γ ⊂ supp(fL), by condition (3), either γ ⊂ supp(fL)∩Wi 0 0 or γ ⊂ (supp(fL) ∩ Uj . If γ ⊂ supp(fL) ∩ Wi , then γ covers the fiber at least ±εL times and hence γ covers the fiber at least |ki| + 1 or −|ki| − 1 times by 0 0 0 condition (1). If γ ⊂ (supp(fL) ∩ Uj then it covers Γj at least |kj| + 1 or −|kj| − 1 times. Consequently, γ is not homotopic to η and is non-contractible as Γj is ON GROWTH RATE AND CONTACT HOMOLOGY 23 not contractible and the fiber is not a torsion element. By Lemma 6.5, αL,λ is hypertight. We now focus on the existence of non-degenerate, hypertight perturbations of 0 0 αL,λ. We may assume that the boundaries of Uj and Wi are tori x = cst with dense 0 Reeb orbits. Choose a small non-degenerate perturbation α of αL,λ that preserves 0 0 the boundaries of Uj and Wi and such that ε is a lower bound for the z-component 0 0 of Rα0 in Wi and for the y-component of Rα0 in Uj. If γ is a periodic Reeb orbit, 0 0 then γ is contained either in a Uj or in a Wi . As in the previous paragraph, γ is non-contractible. Lemma 6.7. Under Hypothesis H, the growth rate of contact homology is (at most) quadratic. Proof. Let α0 be a non-degenerate and hypertight contact form (given for instance by Lemma 6.6). Let αLi,λi be a sequence of contact forms with Li → ∞ such ∗ that Li ∈/ σ(α) and λi ≤ λ(Li) for all i ∈ N . Perturb αLi,λi to obtain a non- degenerate hypertight form α0 (Lemma 6.6). For λ small enough and for small Li,λi i perturbations, the periodic Reeb orbits of α0 with period smaller than L are in Li,λi i bijection with the periodic Reeb orbits of αLi,λi with period smaller than Li and the difference between their period and the period of the associated Rα periodic 1 ∗ orbits is bounded by 2 . Thus, there exists C > 0 such that for all i ∈ N and for all L ≤ Li #Ccyl (M, α0 ) ≤ C#(σ(α) ∩ [0,L + 1]) . ≤L Li,λi Let α0 = f α0 . As the α0 are perturbations of α, there exists D > 0 such Li,λi Li,λi Li,λi that 1 1 < sup fLi,λi (p), < D. D p∈M fLi,λi (p) By invariance of cylindrical contact homology (Corollary 4.10) and by [18, Section 10], there exists C(D) such that, for all L > 0 and for all i, i rk(ψL) ≤ rk(ψC(D)L) where ψi : HCcyl (M, α0 ) → HCcyl(M, α0 ) and ψ : HCcyl (M, α0) → L ≤L Li,λi Li,λi L ≤L HCcyl(M, α0) are the maps defining the direct limit. Hence,

rk(ψL) ≤ C#(σ(α) ∩ [0,C(D)L + 1]) and rk(ψL) exhibits a quadratic growth. 6.4. Holomorphic cylinders and Morse-Bott theory. Let η be a loop that is a multiple of the ﬁber or projects on a multiple of a connected component of

Γ. By Lemma 6.6, there exist L > 0 and λ > 0 such that all the RαL,λ -periodic orbits homotopic to η are non-degenerate, associated to a critical point of fL0 for L0 ≤ L and have period smaller than L. Consider V = W ∪ S U where j j k∈Kj k Kj = {k, Uk is adjacent to Wj}. Then Vj is a trivial circle bundle. Extend the 0 1 trivialization from Proposition 6.2 in Vj ' Sj × S . In these coordinates, α = (f(x) + mg(x)) dy + g(x)dz and the Reeb vector ﬁeld is positively collinear to  0   −g0  . f 0 + mg0

Note that the y-coordinate is negative in Vj \ Wj.

Lemma 6.8. Let u : R × S1 → R × M be a holomorphic cylinder negatively 1 asymptotic to γ ∈ Wj. Then uM (R × S ) ⊂ Wj. 24 ANNE VAUGON

1 Proof. We prove the lemma by contradiction. Assume uM (R × S ) ∩ Ul 6= ∅ for 1 1 l ∈ Kj, then there exists an open interval I such that, in I × S × S ⊂ Ul,

(1) α = f1(x)dθ + g1(x)dz in the trivialization of Vj; 1 1 1 (2) uM (R × S ) ∩ {x} × S × S 6= ∅ for all x ∈ I. By Lemma 5.2, for all x0 ∈ I, −1 1 u u(R × S ) ∩ R × Tx0 1 is a finite union of smooth circles homotopic {∗}×S . For all l ∈ Kj, choose x0 ∈ I and cut R × S1 along the associated circles. Choose the connected component asymptotic to −∞ × S1. Let C denote the oriented boundary of this component and choose a collar neighborhood A = A+ ∪ A− of C as in Lemma 5.3: A± are open annuli in the connected component of R × S1 \ C asymptotic to ±∞ × S1. Let W+ (resp W−) denote the union of the connected components of W such that the Reeb vector field is positively (resp negatively) tangent to the fiber.

If γ ⊂ W+, the line in Tx0 tangent to p = (0, 1) is in the homotopy class of γ. 1 1 Hence (p, Rα) is an oriented basis. By Lemma 5.3, uM (A−) ⊂ (x−, x0) × S × S .

W− uM (A−) uM (A+) W+

x0 x

1 1 Yet uM (A−) ⊂ (x0, x+) × S × S as u is negatively asymptotic to γ. This leads to a contradiction.

If γ ⊂ W−, the line tangent to p = (0, −1) in Tx0 is in the homotopy class of γ 1 1 and (p, Rα) is not an oriented basis. Thus uM (A−) ⊂ (x0, x+) × S × S .

W− uM (A+) uM (A−) W+

x0 x

This leads to a contradiction as u is negatively asymptotic to γ. Lemma 6.9. Let u : R×S1 → R×M a holomorphic cylinder negatively asymptotic 1 1 to γ ∈ Uj. Then uM (R × S ) ⊂ Uj and uM (R × S ) ⊂ supp(fL) 1 1 1 1 Proof. Consider x0 such that γ ∈ Tx0 in the trivialization (− 2 , 2 ) × S × S of Uj. We prove the lemma by contradiction. Thus there exists an open interval 1 1 1 1 1 I ⊂ (− 2 , 2 ) such that α = f(x)dy + g(x)dz and f(R × S ) ∩ {x} × S × S 6= ∅ for all x ∈ I. By Lemma 5.2, for all x1 ∈ I, −1 1 u u(R × S ) ∩ R × Tx1 is a ﬁnite union of smooth circles homotopic to {∗} × S1. Cut R × S1 along these circles and denote by C the oriented boundary of the component asymptotic to 1 −∞ × S . Let p be a vector tangent to Tx0 so that the straight line in Tx0 directed by p is in the homotopy class [γ]. Consider a collar neighborhood A = A+ ∪ A− of 1 C as in Lemma 5.3: A± are open annuli in the connected component of R × S \ C asymptotic to ±∞ × S1. 0 If x1 > x0 then (p, Rα) is not an oriented basis (f is increasing) and uM (A−) ⊂ 1 1 (x1, x+) × S × S by Lemma 5.3. If x1 < x0 then (p, Rα) is an oriented basis and 1 1 uM (A−) ⊂ (x−, x1) × S × S .

uM (A−) uM (A+) uM (A+) uM (A−)

x1 x0 x1 x ON GROWTH RATE AND CONTACT HOMOLOGY 25

This leads to a contradiction as u is negatively asymptotic to γ. Lemma 6.10. For all j = 1 . . . m, there exists a contact closed manifold without boundary (W˜ j, α˜) extending (Wj, α) such that α˜ is of Morse-Bott type. For all i = 1 . . . n, there exists a contact closed manifold without boundary (U˜i, α˜) extending (Ui, α) such that α˜ is of Morse-Bott type. 1 Proof. In the trivialization Wj ' Sj ×S , the contact form is α = β +edz and, near 1 1 1 1 ∂Wj, there exist coordinates (x, y, z) ∈ [0, 1] × S × S such that {1} × S × S ⊂ 0 ∂Wj and α = f(x)dy + edz. Let S be an oriented compact surface such that 0 ∂S and ∂Sj have the same number of connected components. Choose a pairing between these components and glue a neighborhood of each component of ∂Wj to a neighborhood of the associated component of ∂S0 × S1 with the diffeomorphism ˜ ϕ :(x, y, z) 7→ (x, y, z + ky) where k ∈ Z. Let Wj denote the resulting manifold. Near ∂S0 × S1, ∗ ˜ ϕ α = f(x) + ke dy + edz = βk + edz. ˜ 0 For each component, choose k so that eβk is positive on ∂S . There exists a 1-form 0 0 0 ˜ 0 β on S such that edβ > 0 and βk = β near the boundary. The contact form 0 ∗ β + edz extends ϕ α and the induced form α˜ on W˜ j is of Morse-Bott type. 1 On Uj = A × S , the contact form is written α = f(x)dy + g(x)dz. Extend f and g to maps f˜ and g˜ on S1 so that α˜ = f˜(x)dy +g ˜(x)dz is a contact form on T 3. The form α˜ is of Morse-Bott type. Proof of Theorem 1.6. It remains to compute contact homology for η satisfying condition (1) or (2). By Lemma 6.6, there exist L > 0 and λ > 0 such that all the RαL,λ -periodic orbits homotopic to η have a period smaller than L, are non- 0 degenerate and associated to a critical point of fL0 for L ≤ L. ˜ ˜ Extend f L to the contact manifolds Wj and Ui (Lemma 6.10) to get a Morse- Bott perturbation. Let (λl) be a decreasing sequence such that λl ∈ (0, λ] and liml→∞ λl = 0. Choose almost complex structures Jλl adapted to (M, αL,λl ) and ˜ ˜ ˜ Jλl adapted to the union of (Wj, α˜L,λl ) for j = 1 . . . m and (Ui, α˜L,λl ) for i = 1 . . . n so that ˜ (1) Jλl = Jλl on R × Wj and R × Ui; 1 0 ˜ 0 (2) Jλl and Jλl are S -invariant on R × NL for all L ≤ L; ˜ (3) (fL0 , gL0 ) (resp (fL0 , g˜L0 )) is Morse-Smale on SL0 where gL0 (resp g˜L0 ) is the ˜ 0 metric induced by Jλl (resp Jλl ) for all L ≤ L. By Theorems 3.4 and 3.5, for all j = 1 . . . m (resp i = 1 . . . n) and for l big ˜ enough, Jλl -holomorphic cylinders asymptotic periodic Reeb orbits in Wj (resp. Ui) are contained in Wj (resp. Ui) as gradient lines between two points in Sj are contained in Sj. By Lemmas 6.8 and 6.9, for all j = 1 . . . m (resp i = 1 . . . n) and for l big enough, Jλl -holomorphic cylinders asymptotic periodic Reeb orbits in Wj (resp. Ui) are contained in Wj (resp. Ui). Therefore the differential of contact homology is well defined and can be identified with the differential in the Morse-Bott case and thus with the differential in Morse homology. Hence (1) if [η] = [fiber]k with ±k > 0,

[η] M Morse HC∗ (M, α, Q) = H∗ (Wj, (f1, g1), Q) Wj ⊂W± k0 0 (2) if [π(η)] = [Γj] with k 6= 0, [η] M Morse 1 HC∗ (M, α, Q) = H∗ (S , (fL, gL), Q) i∈I 26 ANNE VAUGON where I = {i, [π(Γi)] = [π(Γj)]} (we do not consider the graduation in the identiﬁ- cations). In their study of algebraic torsion in SFT, Latschev and Wendl [43] used similar methods to understand holomorphic curves.

7. Hyperbolicity and exponential growth rate In this section we prove Theorem 1.2. This result hinges on the exponential growth of contact homology for a speciﬁc family of contact structures (Theorem 1.3). The invariance of contact homology leads to the exponential growth of NL(α) for all non-degenerate contact forms. For a general non-degenerate contact form, the proof depends on Hypothesis H. Let M be a 3-manifold which can be cut along a nonempty family of incompressible tori T1,...TN into irreducible manifolds including a hyperbolic component that ﬁbers on the circle. We construct contact forms on each irreducible components and add torsion near the incompressible tori Tk for k = 1 ...N (Section 7.2). We compute the growth rate of contact homology by controlling the holomorphic cylinders that intersect the tori Tk for k = 1 ...N (Section 7.3). The study of periodic orbits and contact homology in the hyperbolic component hinges on properties of periodic points of pseudo-Anosov automorphisms recalled in Section 7.1.

7.1. Periodic points of pseudo-Anosov automorphisms. See [12, 24] for a precise definition of pseudo-Anosov automorphisms. Here, we will only use the properties of pseudo-Anosov automorphisms described in Theorems 7.1, 7.3, 7.4 and 7.5. Let S be a compact orientable surface. An automorphism ψ : S → S is said to be pseudo-Anosov if there exist two measured foliations (F s, µs) and (F u, µu) such that ψ(F u, µu) = (F u, λ−1µu) and ψ(F s, µs) = (F s, λµs) for a positive real number λ. In this section, we assume that S has no boundary. Theorem 7.1 (see [18, Theorem 11.1]). The number of simple k-periodic points of a pseudo-Anosov automorphism on S exhibits an exponential growth with k. This theorem follows from the construction of a Markov partition on S (see [18, Section 11]). Nielsen classes are used to transfer properties of periodic points of a pseudo-Anosov map to properties of periodic points of homotopic diffeomorphisms. Definition 7.2. Let h : S → S be an automorphism. Two fixed points x and y are in the same Nielsen class if there exists a continuous map δ : [0, 1] → S such that δ(0) = x, δ(1) = y and h(δ) is homotopic to δ. Let ht : S → S, t ∈ [0, 1] be a homotopy of automorphism of S. Fixed points x0 of h0 and x1 of h1 are in the same Nielsen class if there exists a continuous map δ : [0, 1] → S such that δ(0) = x, δ(1) = y and h·(δ(·)) is homotopic to δ. One can refer to [25] for more information on Nielsen classes. These definitions extend naturally to periodic points. Two periodic points are in the same Nielsen class of a diffeomorphism h if and only if the induced periodic orbits of the vertical vector field in the mapping torus (S × R)/h are homotopic. Nielsen classes of a pseudo-Anosov automorphisms are very special. Theorem 7.3 (Thurston, Handel [31, Lemma 2.1]). All the periodic points of a pseudo-Anosov automorphism on S are in different Nielsen classes. A k-periodic point x of h : S → S is non-degenerate if 1 is not an eigenvalue k of dh (x). For a non-degenerate k-periodic point, let εhk (x) denote the sign of ON GROWTH RATE AND CONTACT HOMOLOGY 27 det(dhk(x) − Id). If all the periodic points in a given Nielsen class c are non- degenerate, we define X Λhk (c) = εhk (x). x∈c

Theorem 7.4 ([40]). Let h0 and h1 be two homotopic automorphisms of S. Let x0 and x1 be two periodic points of h0 and h1 in the same Nielsen class. If the Nielsen classes c0 of x0 (for h0) and c1 of x1 (for h1) contain only non-degenerate points then Λ k (c0) = Λ k (c1). h0 h1

Theorem 7.5 (see [18, Section 11.1]). Let S1 be a compact surface with boundary obtained from S after removing a ﬁnite number of disjoint open disks. Let h : S1 → S1 be an automorphism such that h = Id in a neighborhood of ∂S1 and homotopic to a pseudo-Anosov automorphism ψ. Extend h to S by the identity and let hˆ denote the resulting automorphism. Then, there exist a branched cover Sˆ of S and a pseudo-Anosov map ψˆ homotopic to hˆ such that the projection of ψˆ is ψ.

7.2. Contact forms on M.

7.2.1. In the hyperbolic component. Let M0 be the hyperbolic component, then M0 is the mapping torus (S × R)/h where (1) S is a compact oriented surface with boundary, (2) h : S → S is a diﬀeomorphism homotopic to a pseudo-Anosov map, (3) h = Id in a neighborhood of ∂S. We use the usual construction on a contact structure on a mapping torus. Choose ∂ cylindrical coordinates (r, θ) in a neighborhood of ∂S so that ∂θ is positively tangent to ∂S. Let β be a 1-form on S such that dβ > 0 and, near ∂S, β = b(r)dθ with b > 0 and b0 > 0. Let F : [0, 1] → [0, 1] be a smooth non-decreasing function such that F = 0 near 0 and F = 1 near 1. On S × [0, 1] consider the contact form

α = (1 − F (t))β + F (t)h∗β + dt where t is the coordinate on [0, 1]. This contact form induces a contact form on M0. The associated contact structure is universally tight.

Lemma 7.6. The Reeb vector ﬁeld is positively transverse to S × {∗} and the ﬁrst return map on S × {0} is homotopic to h.

Proof. If the Reeb vector ﬁeld is tangent to S × {t} in (p, t) then

∗ ιRα ((1 − F (t))dβ + F (t)h dβ)(p, t) = 0 as dα = (1 − F (t))dβ + F (t)h∗dβ + F 0(t)dt ∧ (h∗β − β).

∗ Yet dβ and h dβ are two positive volume forms. Hence Rα is transverse to S ×{t}. It is positively transverse by the boundary condition. The first return map is well ∂ defined and homotopic to h as h is the first return map of ∂t on S × {0} and Rα ∂ and ∂t are homotopic in the space of vector fields transverse to S × {∗}.

In M0, periodic Reeb orbits correspond to periodic points of the ﬁrst return map on S × {0}. Without loss of generality, we may assume that all the periodic points of the ﬁrst return map in the interior of S are non-degenerate. 28 ANNE VAUGON

7.2.2. In non-hyperbolic components. We use the following theorem of Colin and Honda. Theorem 7.7 (Colin-Honda, [17, Théorème 1.3]). Let M be a compact, oriented, irreducible 3-manifold with non-empty boundary such that ∂M is a union of tori. Then there exist an hypertight contact form α on M and a neighborhood T × I of each boundary component (I is an interval) with coordinates (x, y, z) such that α = cos(z)dx − sin(z)dy. In addition there exist arbitrarily small non-degenerate hypertight perturbations of α. The construction in [17] gives the same contact structures as [39] and [16]. With- out loss of generality, all the periodic orbits whose free homotopy classes do not correspond to a class in the boundary are non-degenerate. 7.2.3. Interpolation and torsion. In the previous sections we constructed an hy- SN pertight contact form α on each connected component of M \ k=1 ν(Tk) where ν(Tk) is a neighborhood of Tk. We now glue these components together. Choose k ∈ 1 ...N. There exist coordinates (x, y, z) in a neighborhood Tk ×[a, b] of Tk such that in a neighborhood of T × {a} the contact form is written fa(x)dy + ga(x)dz and in a neighborhood of T × {b} the contact form is written fb(x)dy + gb(x)dz. ∗ Lemma 7.8. For all n ∈ N , there exist fn :[a, b] −→ R and gn :[a, b] −→ R two smooth functions such that

(1) fn extends fa and fb and gn extends ga and gb; (2) α = fn(x)dy + gn(x)dz is a contact form; (3) in coordinates (θ, z), the Reeb vector ﬁeld Rα sweeps out an angle in π 3π 2nπ − , 2nπ + . 2 2 0 0 Proof. The contact condition is fngn − fngn > 0 and the Reeb vector ﬁeld is  0  1 0 Rα = 0 0  −gn  . fngn − fngn 0 fn

The conditions (2) and (3) are equivalent to “the parametrized curve (fn, gn) in 2 π R turns clockwise and its normal vector sweeps out an angle in (2nπ − 2 , 2nπ + 3π 2 2 ]”. We choose a parametric curve in R extending (fa, ga) and (fb, gb) with these properties. ∗ For all n ∈ N , construct a contact form αn on M by extending α by αn = fn(x)dy + gn(x)dz in a neighborhood of T1 and by αn = f1(x)dy + g1(x)dz in a neighborhood of T2,...,TN . 0 Remark 7.9. If {b} × T is in ∂M0, then fb < 0, fb > 0 and gb = 1 near b. If 0 {a} × T is in ∂M0, then fa > 0, fa > 0 and ga = 1 near a (changes in signs are due to the orientation convention of the boundary).

By [15, Théorème 4.2], as contact structures ξn = ker(αn) are universally tight ∗ on each components, (M, ξn) is universally tight for all n ∈ N . In addition, as our construction corresponds to the construction in [16, Section 4], by Theorem [16, Theorem 4.5], there exist inﬁnitely many non-isomorphic ξn. 7.3. Growth rate of contact homology. Lemma 7.10. For all almost complex structures on R × M adapted to the contact form constructed above, there is no holomorphic cylinder u : R × S1 → R × M asymptotic to two periodic Reeb orbits contained in diﬀerent connected components SN of M \ ( k=1 Tk × [a, b]). ON GROWTH RATE AND CONTACT HOMOLOGY 29

Proof. In Tk × [a, b], the contact form is written αn = f(x)dy + g(x)dz and the Reeb vector ﬁeld is  0  1 R = −g0 . αn f 0g − fg0   f 0

It depends only on the x variable and we denote it by Rαn (x). We prove this result by contradiction. If such a u exists then there exists k ∈ 1 ...N such that 1 uM (R × S ) ∩ Tk × {x}= 6 ∅ for all x ∈ [a, b]. By Theorem 2.1 and Proposition 2.2, there exist x0 and x1 in [a, b] such that

(1) Rαn (x0) = −Rαn (x1), −1 1 (2) for all (s, t) ∈ C = u u(R × S ) ∩ (R × {x0, x1} × T ) , we have du(s, t) 6= ∂ 0 and ∂τ ∈/ im(du(s, t)). By Lemma 5.3, C is a ﬁnite union of smooth circles homotopic to {∗} × S1. Cut 1 R × S along these circles and choose a connected component Σ such that uM (Σ) ∩ {x} × T 6= ∅ for all x ∈ [x0, x1]. Then ∂uM (Σ) is a union of two homotopic circles: one in {x0} × T and one in {x1} × T . By positivity of intersection the Reeb vector ﬁeld is positively transverse to these circles in {x0} × T and {x1} × T . This leads to a contradiction as Rαn (x0) = −Rαn (x1).

Let Λ0 be the set of primitive free homotopy classes that correspond to periodic orbits in M0 and do not represent a homotopy class in a torus Tk for k = 1,...,N. All the periodic Reeb orbits with homotopy class in Λ0 are non-degenerate. As there are no contractible periodic orbits, the associated partial contact homology is well defined. There exists C > 0 such that all periodic orbits in M0 associated to a k-periodic point of the first return map h1 have a period smaller than kC. [η] Lemma 7.11. For all η ∈ Λ0, dim(HC∗ (M, αn)) ≥ 1. In addition, if η is associ- [η] [η] ated to k-periodic points, for all L > kC the map HC≤L(M, αn) → HC∗ (M, αn) has a rank greater than 1. η Proof. Choose η ∈ Λ0. Write C∗ = C0 ⊕ C1 where C0 is generated by periodic orbits in M0 homotopic to η and C1 is generated by periodic orbits in M \ M0 homotopic to η. By Lemma 7.10, the differential is written ∂ 0 η . 0 ∗

We prove that dim (ker(∂η)/im(∂η)) ≥ 1. Write C0 = E ⊕ O where E is generated by even periodic orbits and O by odd periodic orbits (as η is primitive all the periodic orbits are good). Then 0 ∂O ∂η = ∂E 0 and ker(∂η)/im(∂η) = ker(∂E)/im(∂O) ⊕ ker(∂O)/im(∂E).

Hence, dim (ker(∂η)/im(∂η)) = {0} if and only if dim(ker(∂E)) = dim(im(∂O)) and dim(ker(∂O)) = dim(im(∂E)). By Section 7.1, there exist a branched cover Sˆ of S and a pseudo-Anosov map ˆ ˆ ˆ ψ such that the lift h1 of h1 is homotopic to ψ. Let c denote the Nielsen class associated to the periodic orbits in M0 homotopic to η and k denote the order of the ˆ associated periodic points. Let cˆ be a Nielsen class of h1 containing a lift of a point in c. As c does not contain points in ∂S, all periodic points in cˆ are non-degenerate and there exists s such that cˆ contains exactly s lifts of each point in c. By Theorems 30 ANNE VAUGON

7.3 and 7.4, Λˆk (ˆc) = Λ ˆk (ˆc) 6= 0. A periodic point of h1 is even if and only if the h1 ψ associated Reeb orbit is even. Therefore, we have Λˆk (ˆc) = s dim(E) − s dim(O) h1 and dim(ker(∂O)) + dim(im(∂O)) 6= dim(ker(∂E)) + dim(im(∂E)). [η] Hence, dim (ker(∂η)/im(∂η)) > 0 and dim(HC∗ (M, αn)) ≥ 1. η For all L > kC, write C≤L = C0 ⊕ C≤L. As the diﬀerential is written ∂η 0 0 ∗ [η] [η] the map HC≤L(M, αn) → HC∗ (M, αn) has a rank greater than 1.

Λ0 Proof of Theorem 1.3. It remains to prove that the growth rate of HC∗ (M, αn) is exponential. By Theorems 7.1, 7.3, 7.4 and 7.5, the number of Nielsen classes associated to periodic points of the first return map h1 grows exponentially. As the number of homotopy classes in tori Tk for k = 1 ...N exhibits a quadratic growth (Lemma 6.4) and by Lemma 7.11, the growth rate of partial cylindrical homology is exponential. Proof of Theorem 1.2. By invariance of the growth rate of partial contact homology (Proposition 4.11), the growth rate of the number of periodic Reeb orbits is exponential if cylindrical contact homology is well defined (Remark 4.5), i.e. if the contact form is non-degenerate and hypertight. p Under Hypothesis H, let αn be a non-degenerate contact form such that ξn = p ker(αn). By Theorem 7.7 and Lemma 6.6 there exists an hypertight non-degenerate 0 contact form αn of ξn. By Theorem 1.3 the growth rate of cylindrical contact p 0 0 homology is exponential. Consider the map A∗(M, αn,J) → A∗(M, αn,J ) given by Theorem 2.5 and the pull back augmentation induced by the trivial augmentation 0 0 on A∗(M, αn,J ) (Proposition 2.11). By invariance of the growth rate of linearized p contact homology (Proposition 4.12), NL(αn) exhibits an exponential growth. References [1] Ian Agol. The virtual Haken conjecture. Documenta Mathematica, 18:1045–1087, 2013. With an appendix by Ian Agol, Daniel Groves, and Jason Manning. [2] Daniel Bennequin. Entrelacement et équations de Pfaff. In IIIe Rencontre de Géométrie du Schnepfenried, volume 1 of Astérisque, pages 87–161, 1983. [3] Laurent Bessières, Gérard Besson, Michel Boileau, Sylvain Maillot, and Joan Porti. Geometri- sation of 3-manifolds, volume 13 of Tracts in Mathematics. European Mathematical Society, 2010. [4] Frédéric Bourgeois. A Morse-Bott approach to Contact Homology. PhD thesis, Stanford Uni- versity, 2002. [5] Frédéric Bourgeois. A Morse-Bott approach to Contact Homology. In Symplectic and Contact Topology : Interactions and Perspectives, volume 35, pages 55–77. 2003. [6] Frédéric Bourgeois. Introduction to Contact Homology. Summer School in Berder, http://www.math.u-psud.fr/∼bourgeois/papers/berder.html, June 2003. [7] Frédéric Bourgeois. A survey of contact homology, volume 49 of CRM Proceedings & Lectures Notes. American Mathematical Society, 2009. [8] Frédéric Bourgeois and Vincent Colin. Homologie de contact des variétés toroïdales. Geometry & Topology, 9:299–313, 2005. [9] Frédéric Bourgeois, Tobias Ekholm, and Yakov Eliashberg. Effect of Legendrian Surgery. Geometry & Topology, 16:301–389, 2012. [10] Frédéric Bourgeois, Yakov Eliashberg, Helmut Hofer, Kris Wysocki, and Eduard Zehnder. Compactness results in symplectic field theory. Geometry & Topology, 7:799–888, 2003. [11] Frédéric Bourgeois and Klaus Mohnke. Coherent orientations in symplectic field theory. Math- ematische Zeitschrift, 248(1):123–146, 2003. [12] Andrew J. Casson and Steven A. Bleiler. Automorphisms of Surfaces after Nielsen and Thurston, volume 9 of London Mathematical Society Student Texts. Cambrige University Press, 1988. ON GROWTH RATE AND CONTACT HOMOLOGY 31

[13] Yuri Chekanov. Differential algebra of Lengendrian links. Inventiones Mathematicae, 150:441–483, 2002. [14] Kai Cieliebak and Janko Latschev. The role of string topology in symplectic field theory. In New perspectives and challenges in symplectic field theory, volume 49 of CRM Proceedings & Lectures Notes, pages 113–146, 2009. [15] Vincent Colin. Recollement de variétés de contact tendues. Bulletin de la Société mathéma- tique de France, 127(1):43–69, 1999. [16] Vincent Colin. Une infinité de structures de contact tendues sur les variétés toroïdales. Com- mentarii Mathematici Helvetici, 76(2):353–372, 2001. [17] Vincent Colin and Ko Honda. Constructions contrôlées de champs de Reeb et applications. Geometry & Topology, 9:2193–2226, 2005. [18] Vincent Colin and Ko Honda. Reeb Vector Fields and Open Book Decompositions. Journal of European Mathematical Society, 15:443–507, 2013. [19] Pierre de la Harpe. Topics in Geometric Group Theory. The University of Chicago Press, 2000. [20] Dragomir Dragnev. Fredholm theory and transversality for non compact pseudoholomorphic curves in symplectisations. Communications on pure and applied mathematics, 57:726–763, 2004. [21] Yakov Eliashberg. Classification of overtwisted contact structures on 3-manifolds. Inventiones Mathematicae, 98(3):623–637, 1989. [22] Yakov Eliashberg, Alexander Givental, and Helmut Hofer. Introduction to Symplectic Field Theory. Geometric and Functional Analysis (GAFA), pages 560–673, 2000. Special volume, Part II. [23] Oliver Fabert, Joel Fish, Roman Golovko, and Katrin Wehrheim. Polyfolds: a first and second look. arXiv: 1210.6670, 2012. [24] Albert Fathi, François Laudenbach, and Valentin Poénaru. Travaux de Thurston sur les surfaces, volume 66-67 of Astérisque. Société mathématique de France, seconde edition, 1991. Séminaire Orsay. [25] Alexander Fel’shtyn. Dynamical Zeta Functions, Nielsen Theory and Reidemeister torsion, volume 147 of Memoirs of the American Mathematical Society. American Mathematical So- ciety, Providence, RI, 2000. [26] Kenji Fukaya and Kaoru Ono. Arnold conjecture and Gromov-Witten invariants. Topology, 38:933–1048, 1999. [27] Hansjörg Geiges. An Introduction to Contact Topology. Cambrige University Press, 2008. [28] Emmanuel Giroux. Structures de contact en dimension trois et bifurcations des feuilletages de surfaces. Inventiones Mathematicae, 141:615–689, 2000. [29] Emmanuel Giroux. Structures de contact sur les variétés fibrées en cercles au-dessus d’une surface. Commentarii Mathematici Helvetici, 76:218–262, 2001. [30] Mikhaïl Gromov. Pseudo-holomorphic curves in symplectic manifolds. Inventiones Mathe- maticae, 82:307–347, 1985. [31] M Handel. Global shadowing of pseudo-Anosov homeomorphism. Ergodic Theory and Dy- namical Systems, 5:373–377, 1985. [32] Helmut Hofer. Pseudoholomorphic curves in symplectization with applications to the Wein- stein conjecture in dimension three. Inventiones mathematicae, 114:515–563, 1993. [33] Helmut Hofer. Polyfolds and a General Fredholm Theory. arXiv:0809.3753, September 2008. To appear in Proceedings of the 2008 Clay research conference. [34] Helmut Hofer, Kris Wysocki, and Eduard Zehnder. Applications of Polyfold Theory II : The Polyfolds of Symplectic Field Theory. In preparation. [35] Helmut Hofer, Kris Wysocki, and Eduard Zehnder. Properties of Pseudoholomorphic Curves in Symplectisations II: Embedding contril and Algebraic invariants. Geometric and Func- tional Analysis (GAFA), 5(2), 1995. [36] Helmut Hofer, Kris Wysocki, and Eduard Zehnder. Properties of Pseudoholomorphic Curves in Symplectisations I: Asymptotics. Annales de l’institut Henri Poincaré, pages 337–379, 1996. [37] Helmut Hofer, Kris Wysocki, and Eduard Zehnder. Applications to Polyfold Theory I : The Polyfolds of Gromov-Witten Theory. arXiv:1107.2097, July 2011. [38] Ko Honda. On the classification of tight contact structures II. Journal of Differential Geom- etry, 55(1), 2000. [39] Ko Honda, Will Kazez, and Gordana Matić. Tight contact structures and taut foliations. Geometry & Topology, 4:219–242, 2000. [40] BoJu Jiang. Lectures on Nielsen fixed point theory, volume 14 of Contemporary Mathematics. American Mathematical Society, 1983. 32 ANNE VAUGON

[41] Anatole Katok and Boris Hasselblatt. Introduction to the modern theory of dynamical systems, volume 54 of Encyclopedia of mathematics and its applications. Cambrige University Press, 1995. [42] Wilhelm Klingenberg. Riemannian Geometry, volume 1 of Studies en mathematics. Gruyter and Co, Berlin, 1982. [43] Janko Latschev and Chris Wendl. Algebraic Torsion in Contact Manifolds. Geometric and Functional Analysis (GAFA), 21(5):1144–1195, 2011. With an appendix by Michael Hutch- ings. [44] François Laudenbach. Symplectic geometry and Floer homology. Ensaios Mathematicos, 7:1– 50, 2004. [45] Gang Liu and Gang Tian. Floer homology and Arnold conjecture. Journal of Differential Geometry, 49:1–74, 1998. [46] Dusa McDuff. Singularities and positivity of intersections of J-holomorphic curves. In Michèle Audin and Jacques Lafontaine, editors, Holomorphic curves in symplectic geometry, volume 117 of Progress in Mathematics. Birkhäuser, 1994. With an appendix by Gang Liu. [47] Dusa McDuff and Dietmar Salamon. J-holomorphic Curves and Symplectic Topology, volume 52 of Colloquium Publications. American Mathematical Society, Providence, RI, 2004. [48] Mark McLean. The growth rate of symplectic homology and affine varieties. Geometric and Functional Analysis (GAFA), 22(2):369–442, 2012. [49] John Milnor. A note on curvature and fundamental group. Journal of Differential Geometry, 2:1–7, 1968. [50] Paul Seidel. A biased view of symplectic cohomology. Current Developments in Mathematics, 2006:211–253, 2008. [51] Clifford Henry Taubes. The Seiberg-Witten equations and the Weinstein conjecture. Geom- etry & Topology, 11:2117–2202, 2007. [52] William Thurston. Three-dimensional manifolds, Kleinian groups and hyperbolic geometry. Bulletin (New Series) of the american mathematical society, 6(3):357–391, 1982. [53] William Thurston. Minimal stretch maps between hyperbolic surfaces, 1998. [54] Ilya Ustilovsky. Infinitely many contact structures on S4m+1. International Mathematics Research Notices, 14:781–791, 1999. [55] Anne Vaugon. Étude dynamique des champs de Reeb et propriétés de croissance de l’homologie de contact. Thèse de doctorat, Université de Nantes, December 2011. [56] Mei Lin Yau. Vanishing of the contact homology of overtwisted contact 3–manifolds. Bulletin of the Institute of Mathematics Academia Sinica (New Series), 1(2):211–229, 2006. With an appendix by Yakov Eliashberg.

Anne Vaugon, Laboratoire de Mathématiques Jean Leray, 2, rue de la Houssinière, BP 92208, 44322 Nantes Cedex 3, France E-mail address: [email protected]